Provided in the present disclosure are an operator optimization method and apparatus, and a storage medium and an electronic device. The operator optimization method provided in the present disclosure comprises: acquiring a target neural network model, and determining a computational graph of the target neural network model; for each operator in the computational graph, determining a search space including all feasible solutions for the operator; selecting, from the search space, several feasible solutions as candidate solutions, determining evaluation values of the candidate solutions, and using a candidate solution having the highest evaluation value as a tentative solution; determining a running time during which target hardware runs the tentative solution, and increasing the number of iterations; when the running time is less than the current optimal time or there is no current optimal time, determining the running time to be the current optimal time, and determining the tentative solution to be the current optimal solution; when the number of iterations is less than a specified number of times, reselecting, from the search space of the operator, a specified number of candidate solutions that have not been selected; and when the number of iterations is not less than the specified number of times, determining the current optimal solution to be an optimal solution for the operator.
A point cloud matching method and apparatus, an electronic apparatus, and a storage medium. The point cloud matching method comprises: dividing a point cloud to be matched into a plurality of point cloud blocks, inputting the plurality of point cloud blocks into a pre-trained point cloud feature encoding module, and obtaining a feature vector of the plurality of point cloud blocks; on the basis of the feature vector of the plurality of point cloud blocks, obtaining a global descriptor vector of said point cloud, matching the global descriptor vector of said point cloud with a global descriptor vector of a point cloud frame in a preset historical database, and determining a point cloud frame in a historical database within a preset matching threshold range to be a point cloud matching result.
G06T 7/33 - Détermination des paramètres de transformation pour l'alignement des images, c.-à-d. recalage des images utilisant des procédés basés sur les caractéristiques
3.
CONVERGED DATA EXCHANGE METHOD AND TIME-SENSITIVE NETWORK SWITCH
A converged data exchange method and a time-sensitive network switch are provided in the present disclosure. The network switch device performs sensitivity identification for the data to be forwarded. For the switching transmission of non-time-sensitive data, the push port queue and the pop port queue perform cut-through switching for the data, which reduces the residence time of the data packet in the switch device. For time-sensitive data, the push port queue and pop port queue forward the data according to the corresponding priority scheduling policies.
H04L 47/6275 - Ordonnancement des files d’attente caractérisé par des critères d’ordonnancement pour des créneaux de service ou des commandes de service basé sur la priorité
H04L 47/12 - Prévention de la congestionRécupération de la congestion
H04L 47/2425 - Trafic caractérisé par des attributs spécifiques, p. ex. la priorité ou QoS pour la prise en charge de spécifications de services, p. ex. SLA
4.
MULTIMODAL KEYWORD SPOTTING METHOD AND DEVICE BASED ON AUDIO AND VIDEO
A multimodal keyword spotting method and device based on audio and a video. The method comprises: acquiring an image sequence and an audio sequence; performing lip detection processing on the image sequence, extracting a detected lip part image, and using a lip feature extraction neural network to process an extracted lip image sequence to obtain an image feature; extracting an audio feature of the audio sequence, and using an audio feature processing neural network to process the extracted audio feature to obtain a high-dimensional audio feature; carrying out feature fusion on the image feature and the high-dimensional audio feature; using a multi-modal feature processing neural network to process an audio and video feature obtained by fusion, to obtain a multi-modal high-dimensional feature; fusing the image feature, the high-dimensional audio feature and the multi-modal high-dimensional feature into a mixed high-dimensional feature; and on the basis of the mixed high-dimensional feature, using a keyword spotting classifier to determine whether spotting needs to be performed.
G10L 15/02 - Extraction de caractéristiques pour la reconnaissance de la paroleSélection d'unités de reconnaissance
G10L 15/06 - Création de gabarits de référenceEntraînement des systèmes de reconnaissance de la parole, p. ex. adaptation aux caractéristiques de la voix du locuteur
G10L 15/16 - Classement ou recherche de la parole utilisant des réseaux neuronaux artificiels
G10L 15/26 - Systèmes de synthèse de texte à partir de la parole
G10L 15/25 - Reconnaissance de la parole utilisant des caractéristiques non acoustiques utilisant la position des lèvres, le mouvement des lèvres ou l’analyse du visage
5.
HUMANOID ROBOT THIGH, HUMANOID ROBOT, AND MANUFACTURING METHOD
Disclosed in the present invention are a humanoid robot thigh, a humanoid robot, and a manufacturing method. The thigh is of an integrally-formed structure, and all components are integrally formed by means of a selective laser melting process without a mechanical connection joint. The thigh comprises a skeleton, skin, and a first equal-density lattice filled between the skeleton and the skin; the skeleton is in a configuration of "local housing thickening + second equal-density lattice"; the skin is a thin housing with equal thickness; the skeleton and the skin are combined into a whole; the first equal-density lattice is added between the skeleton and the skin; and the skeleton comprises a housing, a first reinforcing plate, a second reinforcing plate, and a second equal-density lattice with which the internal space of the skeleton is filled. According to the present invention, the problem of vibration of a plastic appearance part during robot movement due to mechanical connection is effectively avoided, and the assembly time of a leg part is saved. In addition, the configuration design is relatively simple, and the total weight of the leg part is effectively reduced while the rigidity meets the use requirement, thereby prolonging the service life of the thigh part.
B62D 57/032 - Véhicules caractérisés par des moyens de propulsion ou de prise avec le sol autres que les roues ou les chenilles, seuls ou en complément aux roues ou aux chenilles avec moyens de propulsion en prise avec le sol, p. ex. par jambes mécaniques avec une base de support et des jambes soulevées alternativement ou dans un ordre déterminéVéhicules caractérisés par des moyens de propulsion ou de prise avec le sol autres que les roues ou les chenilles, seuls ou en complément aux roues ou aux chenilles avec moyens de propulsion en prise avec le sol, p. ex. par jambes mécaniques avec des pieds ou des patins soulevés alternativement ou dans un ordre déterminé
A point cloud matching method and apparatus, an electronic apparatus, and a storage medium are provided. The point cloud matching method includes: dividing a to-be-matched point cloud into a plurality of matched point cloud patches, and inputting the plurality of matched point cloud patches to a pre-trained point cloud feature coding module to obtain feature vectors of the plurality of matched point cloud patches; and acquiring a global description vector of the to-be-matched point cloud according to the feature vectors of the plurality of matched point cloud patches, matching the global description vector of the to-be-matched point cloud with global description vectors of point cloud frames in a preset historical database, and determining a point cloud frame in the historical database within a preset matching threshold range to be a point cloud matching result.
A method of signaling interaction in 5G space-ground integrated heterogeneous network architecture is provided, which enables publicly available data for trusted satellite discovery to be registered to the 5G core network based on the function framework of the 5G core network. When searching for satellites, a gNB can utilize the publicly available data of the trusted satellite registered in the 5G core network to speed up satellite searching, improve satellite link services, and filter out untrusted satellites. By the method provided by the present disclosure, it can effectively improve the satellite search speed in the 5G space-ground integrated architecture, improve satellite link availability, reduce authentication complexity of satellites in ground networks in heterogeneous network architecture.
The present disclosure provides a method and device for performing signal communication based on a levitated particle. In one example, the method includes: preparing a levitated state of the particle; regulating and measuring a net charge quantity carried by the levitated particle; calibrating electromagnetic response characteristics of the levitated particle; applying an electromagnetic communication signal; obtaining and demodulating the electromagnetic communication signal. In an example, the device includes: a levitation trapper; a charge measure-regulator; an electromagnetic response calibrator, configured to obtain, in advance, a background noise and an electromagnetic response transfer function of the levitated particle; a communication signal detect-demodulator, configured to detect a motion response of the levitated particle under an electromagnetic communication signal; based on the background noise and the electromagnetic response transfer function of the levitated particle, recover the applied electromagnetic communication signal from the detected motion response, and demodulate symbols of the electromagnetic communication signal.
An incremental compiling method and an incremental compiling system based on a heterogeneous device. The method comprises: a user using P4 to develop a network program, checking the syntax and semantics of the network program and converting into an intermediate representation; allocating intermediate representation files according to constraints such as the functions and resources of the heterogeneous device; then, transforming the network program into different rule expressions of the heterogeneous device according to the constraints of the heterogeneous device, flow table entries and the like, and loading the on different rule expressions to a target device; when the user needs to deploy network functions incrementally, the user only needs to write new network programs and regenerate rule expressions; by comparing and analyzing the old and new rule expressions, a rule expression that can be loaded on heterogeneous device is formed, and the incremental deployment of network functions is realized.
Provided in the present disclosure are a method and apparatus for dynamic facial image alignment, and a device and a readable storage medium. An initial mesh template of each scale is optimized according to a first facial image, such that an optimized intermediate mesh template is obtained; a first rendered image is generated according to an intermediate network template and the first facial image; by taking the minimization of the difference between a depth map corresponding to a point cloud of a second facial image and a depth map of the intermediate mesh template, the minimization of the difference between a normal map corresponding to the point cloud of the second facial image and a normal map of the intermediate mesh template and the minimization of the difference between the first rendered image and the second facial image as optimization objectives, the intermediate mesh template is optimized, such that a target mesh template is obtained; and a second rendered image is then obtained on the basis of the target mesh template.
A multi-modal network system (100) and a multi-modal network operation method. The multi-modal network system (100) sequentially comprises from top to bottom: an application layer (11) used for providing network applications corresponding to network service requirements; a service layer (12) used for determining network capability requirements for implementing the network applications; a modal layer (13) used for providing corresponding network modalities on the basis of the network capability requirements; and an environment layer (14) used for providing network infrastructures supporting the operation of the network modalities, wherein the network infrastructures are used for bearing and transmitting packets corresponding to the network applications, and the packets are generated, encapsulated, decapsulated and routed and forwarded on the basis of the network modalities corresponding to the network applications.
H04L 67/12 - Protocoles spécialement adaptés aux environnements propriétaires ou de mise en réseau pour un usage spécial, p. ex. les réseaux médicaux, les réseaux de capteurs, les réseaux dans les véhicules ou les réseaux de mesure à distance
12.
COREFERENCE RESOLUTION METHOD AND APPARATUS BASED ON REFERENCE TO EXTERNAL KNOWLEDGE
Disclosed in the present invention are a coreference resolution method and apparatus based on reference to external knowledge. The method comprises: generating training data, and establishing and training a mention recognition model and a relation classification model; firstly, inputting a sentence to train the mention recognition model, wherein the model marks mentions in the sentence; and inputting the sentence in which two or three mentions are designated to be spliced with knowledge corresponding to the mentions, and training the relation classification model to determine whether a coreference relation exists among the designated mentions, and marking the mentions having a coreference relation. After model training is completed, the models are used for coreference resolution. According to the method of the present invention, in the process of referring to external knowledge for coreference resolution, semantic information of the whole sentence is considered. In the training process of the relation classification model, the model is trained to determine whether a coreference relation exists among mentions, the model is trained to mark the mentions having a coreference relation, and the input comprises designated three mentions. Thus, the training method enables the model to have deeper understanding for the mentions and the coreference relation, so that the model has stronger coreference resolution capability.
An information recommendation method, an apparatus, a device, and a medium based on embedding table compression are provided. The method includes: determining, based on a preset compression ratio, to-be-compressed features and non-compressed features in a to-be-compressed embedding table of a recommendation model, generating a similarity index matrix based on a similarity between the to-be-compressed features and the uncompressed features; generating an index dictionary based on the similarity index matrix; substituting a first feature mapping dictionary based on the index dictionary to generate a second feature mapping dictionary, wherein the first feature mapping dictionary is generated based on a data set; and acquiring to-be-recommended data, replacing features in the to-be-recommended data according to the second feature mapping dictionary, inputting replaced features into the recommendation model, and outputting a prediction result.
A clinical risk prediction system oriented to data distribution drift detection and self-adaptation, comprising a central server comprising a first drift detection module and a model aggregation module, and nodes comprising a data acquisition module configured to acquire patient clinical diagnosis and treatment data, a second drift detection module and a model updating module. The first and second drift detection module determine whether the patient clinical diagnosis and treatment data distribution has drifted according to whether the new/old patient clinical diagnosis and treatment data set comes from the same data distribution. When the data distribution has drifted, a local clinical risk prediction model is trained, and its parameters are uploaded to the central server and aggregated to obtain an updated model, which is issued to each node for deployment. The new patient clinical diagnosis and treatment data is input into the updated model to obtain a clinical risk prediction result.
G16H 50/30 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour le calcul des indices de santéTIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour l’évaluation des risques pour la santé d’une personne
G06F 18/2413 - Techniques de classification relatives au modèle de classification, p. ex. approches paramétriques ou non paramétriques basées sur les distances des motifs d'entraînement ou de référence
G16H 50/20 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour le diagnostic assisté par ordinateur, p. ex. basé sur des systèmes experts médicaux
15.
POLYMORPHIC NETWORK SYSTEM AND POLYMORPHIC NETWORK OPERATION METHOD
A polymorphic network system and a polymorphic network operation method are provided. From top to bottom, the polymorphic network system sequentially includes: an application layer configured to provide network applications corresponding to network service requirements; a service layer configured to determine network capability requirements for implementing the network applications; a mode layer configured to provide corresponding network modes based on the network capability requirements; and an environment layer configured to provide network infrastructure that is capable of supporting operation of the network modes. The network infrastructure is configured to load and transmit messages corresponding to the network applications. The messages are capable of being generated, encapsulated, decapsulated, and routed and forwarded based on the network modes corresponding to the network applications.
An information recommendation method and apparatus based on embedding table compression, a device, and a medium. The method comprises: on the basis of a preset compression ratio, determining a feature to be compressed and a non-compressed feature in an embedding table to be compressed of a recommendation model; generating a similarity index matrix on the basis of the similarity between the feature to be compressed and the non-compressed feature; generating an index dictionary on the basis of the similarity index matrix; replacing a first feature mapping dictionary on the basis of the index dictionary, and generating a second feature mapping dictionary; and acquiring data to be recommended, replacing a feature in said data on the basis of the second feature mapping dictionary, inputting the replaced feature into the recommendation model, and outputting a prediction result.
Disclosed in the present invention are an internal and external field combined polymorphic network test environment construction method and apparatus. The method comprises: constructing an internal field test environment and an external field test environment which are based on programmable switches; constructing a functional level comprising four modules, i.e., network access, a network facility, network management and control, and a service application; and constructing operation logic, comprising monitoring a network state in real time, analyzing user experiment requirements, issuing resource configuration, operating a user experiment, and updating network resources. The present invention provides a flexible programmable forwarding capability, can support new user-defined protocols while supporting a mainstream IP protocol, and has good expansibility. In addition, the present invention can simultaneously support a laboratory environment and a real network operation environment, thereby being capable of providing a real network test environment for industry users to test new functions and the like, and having good actual application value.
Disclosed in the present invention are a method and device for service validation under a heterogeneous network architecture, for use in a fifth-generation communication system. The method comprises: first deploying a user validation function for a control plane network element; then a control plane newly adding a user validation information reporting trigger condition in reporting trigger conditions of a flow detection rule and a flow control rule, and issuing the flow detection rule; a user plane successfully loading the reporting trigger condition, and upon triggering the newly added reporting trigger condition, inserting user validation information into user validation information newly added in a PFCP session report request; the control plane receiving the PFCP session report request of the user plane, detecting whether the report type is a user validation information report, and performing verification; and if a verification result indicates that the verification is successful, allowing a user session service, or otherwise, triggering a session release procedure. According to the present invention, a data network end can effectively improve, without performing security protection upgrade, the level of preventing pseudo users and illegitimate services from attacking a service server, and effective technical supplementation is carried out on a satellite-terrestrial integrated security technology architecture.
The present disclosure relates to communications technologies, and provides a single polarized ME dipole and a basic radiator of a phased array antenna for millimeter wave or above frequency band. The radiator is an “array of the ME dipoles” with higher dielectric constant, and can be manufactured with the multilayer PCB or LTCC process. The ME dipole comprises an electric dipole made by two parts of a first metal sheet located on a top substrate layer, a magnetic dipole made by first vias connected to the first metal sheet and a third metal sheet located on a bottom substrate layer, a balun made by an other part of the first metal sheet and a third via, and a reflecting part made by a part of the third metal sheet. The substrate layers are stacked in a multilayer structure to facilitate arranging the metal sheets and the vias.
H01Q 9/28 - Éléments coniques, cylindriques, en cage, en ruban, en treillis ou éléments analogues ayant une surface de rayonnement étendueÉléments comportant deux surfaces coniques ayant le même axe et opposées par leurs sommets et alimentés par des lignes de transmission à deux conducteurs
H01Q 1/38 - Forme structurale pour éléments rayonnants, p. ex. cône, spirale, parapluie formés par une couche conductrice sur un support isolant
Disclosed in the present invention are a pipeline security improvement method and apparatus for a space-ground network architecture. The method comprises: pre-registering base station configuration information into a network repository function network element of a 5G core network; after receiving a relay pipeline connection establishment request initiated by a base station, by means of a network exposure function network element, a satellite sending a network function search request to the network repository function network element; after the network exposure function network element receives a network function discovery request response of the base station, on the basis of the presence or absence of the base station registration information, setting a corresponding verification result of the base station and responding to a network element verification request; and finally, the satellite checking a verification result signal element carried by a received response, so as to determine the validity of the currently accessed base station. The present method provides a data pipeline architecture and interaction method for network function sharing and publicly available information among heterogeneous systems, reduces deployment costs, ensures efficiency and universality, and effectively replenishes signaling interaction methods and signal element design which are required for relay satellites identifying the security of pseudo base stations under 5G space-ground integrated architectures.
Disclosed in the present invention is a terahertz kinetic inductance bolometer, including a superconducting thin film layer, a terahertz antenna, a cutoff layer and a Si substrate, wherein the superconducting thin film layer and the terahertz antenna are respectively deposited on the cutoff layer, and the cutoff layer is deposited on the Si substrate; the superconducting thin film layer includes a superconducting feeder line, an inter-digital capacitor and an inductor coil; the inter-digital capacitor is connected with the inductor coil in parallel to form an oscillation circuit; the terahertz antenna is adjacent to the inductor coil and is used to convert a received terahertz signal into heat so that the inductor coil produces an inductance change; a resonance frequency in the inter-digital capacitor changes through the inductance change; and the superconducting feeder line receives the varying resonance frequency, through which an light intensity of the terahertz signal can be obtained to complete the detection of the terahertz signal. The terahertz kinetic inductance bolometer can detect the terahertz signal accurately and is less affected by the temperature. The present invention also provides a preparation method of the terahertz kinetic inductance bolometer and a terahertz detection system.
A data storage system includes: a server, a client and a control end; the control end is configured to generate a configuration file, and send the configuration file to the client and the server; the client is configured to generate an encapsulation rule based on the configuration file, generate a storage request, perform encapsulation on the storage request to obtain a message packet, and send the message packet to the server; the server is configured to generate an extraction unit and an action unit based on the configuration file, analyze the message packet to obtain the target information, write the target information into each extraction unit, read action information and determine an action unit matching the action information as a target action unit, and execute the storage actions corresponding to the target action unit to store byte stream data of the target information.
H04L 41/082 - Réglages de configuration caractérisés par les conditions déclenchant un changement de paramètres la condition étant des mises à jour ou des mises à niveau des fonctionnalités réseau
H04L 41/0894 - Gestion de la configuration du réseau basée sur des règles
H04L 67/1097 - Protocoles dans lesquels une application est distribuée parmi les nœuds du réseau pour le stockage distribué de données dans des réseaux, p. ex. dispositions de transport pour le système de fichiers réseau [NFS], réseaux de stockage [SAN] ou stockage en réseau [NAS]
23.
DATA STORAGE SYSTEM AND METHOD, AND STORAGE MEDIUM AND ELECTRONIC DEVICE
The present application provides a data storage system and method, and a storage medium and an electronic device. The data storage system comprises a server, a client and a control end, wherein the control end is used for generating a configuration file according to programming information, and sending the configuration file to the client and the server; the client is used for generating an encapsulation rule according to the configuration file, generating a storage request, encapsulating the storage request on the basis of the encapsulation rule to obtain a message packet, and sending the message packet to the server; and the server is used for generating extraction units and action units according to the configuration file, parsing the message packet, writing target information into each extraction unit, reading action information from a specified extraction unit, determining an action unit matching the action information to serve as a target action unit, and executing a storage action corresponding to the target action unit, so as to store byte stream data of the target information.
H04L 67/1097 - Protocoles dans lesquels une application est distribuée parmi les nœuds du réseau pour le stockage distribué de données dans des réseaux, p. ex. dispositions de transport pour le système de fichiers réseau [NFS], réseaux de stockage [SAN] ou stockage en réseau [NAS]
A model training method and apparatus, and a molecular property information prediction method and apparatus. The model training method comprises: acquiring data of a specified protein degradation targeting chimera molecule; on the basis of the data, constructing three-dimensional molecular map data of the specified protein degradation targeting chimera molecule; inputting the three-dimensional molecular map data of the specified protein degradation targeting chimera molecule into a prediction model to be trained, so that said prediction model predicts the molecular property information of the specified protein degradation targeting chimera molecule; and training said prediction model on the basis of a difference between the predicted molecular property information and actual molecular property information corresponding to the specified protein degradation targeting chimera molecule.
G16B 15/30 - Ciblage de médicament à l’aide de données structurellesPrévision d’amarrage ou de liaison moléculaire
G16B 15/20 - Repliement de protéines ou de domaines
G16B 40/00 - TIC spécialement adaptées aux biostatistiquesTIC spécialement adaptées à l’apprentissage automatique ou à l’exploration de données liées à la bio-informatique, p. ex. extraction de connaissances ou détection de motifs
25.
HEAT-DRIVEN FULL-SEA-DEPTH LENS WIPING APPARATUS AND METHOD
A heat-driven full-sea-depth lens (12) wiping apparatus and method. The full-sea-depth lens (12) wiping apparatus comprises: a wiper hingedly mounted on a housing; a shape memory alloy wire (7) mounted on the housing and configured to be connected to and drive the wiper to perform wiping motion; a fixing member used for performing locating arrangement on the shape memory alloy wire (7); and a watertight connector (11) electrically connected to the shape memory alloy wire (12). The shape memory alloy wire (7) is electrified and heated to extend and shrink, so as to pull the wiper to rotate and wipe the lens. The present invention has the advantages such as full sea depth, low price, small noise, and high reliability.
Disclosed in the present description are a heterogeneous chip task scheduling method and apparatus based on sequence generation. The method comprises: for each task to be scheduled among a plurality of tasks to be scheduled of a scheduling node, determining execution time data corresponding to said task on the basis of task information of said task; then, determining idle moments respectively corresponding to a plurality of chips; in response to a scheduling request, generating a scheduling sequence according to the respective task information of the plurality of tasks to be scheduled, the idle moments respectively corresponding to the plurality of chips, and the execution time data respectively corresponding to the plurality of tasks to be scheduled; and according to the scheduling sequence, scheduling each task to be scheduled in the scheduling sequence to a corresponding chip for execution. In this way, when there is a chip in an idle state in a computing cluster that includes heterogeneous chips, a task matching the chip in the idle state is allocated thereto as much as possible, thereby ensuring the task execution efficiency.
Disclosed in the present invention are a drug-based cancer inhibition sensitivity prediction method and device based on a graph neural network. The method comprises: acquiring a candidate drug and cancer cell line original data; extracting drug information features of the candidate drug, and expanding the drug information features; constructing a drug molecule feature vector according to the drug information features, and constructing a drug adjacency matrix, a node feature matrix, and an edge feature matrix by taking atoms of a drug as nodes and chemical bonds as edges; constructing and iterating a graph neural network model to obtain a drug graph feature; obtaining a drug characterization after the drug graph feature and the drug molecule feature vector are aggregated; acquiring and storing the expression quantities of gene expression, gene mutation, gene methylation and gene copy number, in a cancer cell line, of a gene sequence, and extracting a cancer cell line characterization; and performing feature fusion on the drug characterization and the cancer cell line characterization to obtain a drug-cancer cell line instance pair, inputting the instance pair into a drug-based cancer inhibition sensitivity prediction model, and predicting to obtain an IC50 value of the candidate drug in a cancer cell line environment.
Disclosed in the present invention is a drug property prediction method based on a graph neural network using vector quantization, comprising: acquiring a drug molecule original graph; constructing a graph encoder, and encoding the drug molecule original graph into latent variable features; constructing a codebook; calculating the Euclidean distances between the latent variable feature corresponding to each node and vectors in the codebook, and replacing the latent variable feature with the nearest neighbor vector, so as to obtain vectorized latent variable features; constructing a graph decoder, and decoding the vectorized latent variable features, so as to obtain a drug molecule enhanced graph; constructing a drug molecule original graph-enhanced graph instance pair, and inputting same into a drug molecule graph comparison network, so as to obtain drug molecule graph instance pair features; constructing a loss function, and co-training the drug molecule enhanced graph and the drug molecule graph comparison network; and inputting graph features obtained after graph feature encoding of a drug molecule graph into the trained graph comparison network for prediction, so as to obtain a drug property prediction result.
G16C 20/50 - Conception moléculaire, p. ex. de médicaments
G16C 20/90 - Langages de programmationArchitectures informatiquesSystèmes de bases de donnéesStockage de données
G16H 70/40 - TIC spécialement adaptées au maniement ou au traitement de références médicales concernant des médicaments, p. ex. leurs effets secondaires ou leur usage prévu
G06N 3/04 - Architecture, p. ex. topologie d'interconnexion
29.
SYSTEM FOR CLASSIFYING WORKING MEMORY TASK MAGNETOENCEPHALOGRAPHY BASED ON MACHINE LEARNING
A system for classifying working memory task magnetoencephalography based on machine learning, including: the magnetoencephalography data acquisition module configured to acquire magnetoencephalography data of a subject in different working memory task states; the magnetoencephalography data preprocessing module configured to control the quality of magnetoencephalography data in different working memory tasks and separate noises and artifacts; the magnetoencephalography source reconstruction module configured for sensor signal analysis and source reconstruction analysis for the data processed by the magnetoencephalography data preprocessing module; and the machine learning classification module is configured to classify the working memory tasks to which the subjects belong by taking power time series as features. The present disclosure integrates the complete analysis pipeline from preprocessing to source reconstruction of the working memory magnetoencephalography data, classifies the working memory task magnetoencephalography data, and is of great significance to the study of working memory decoding and brain memory related mechanisms.
A61B 5/245 - Détection de champs biomagnétiques, p. ex. de champs magnétiques produits par des courants bioélectriques spécialement adaptée aux signaux magnétoencéphalographiques [MEG]
A61B 5/00 - Mesure servant à établir un diagnostic Identification des individus
30.
WHITE-BOX ENCRYPTION METHOD AND SYSTEM BASED ON NEURAL NETWORK
A white-box encryption method based on a neural network. The method comprises a confusion technology and white-box AES encryption and decryption processes. Core steps of an AES solution for a fixed key are summarized into a plurality of mapping tables, and the neural network replaces encryption and decryption operations by means of over-fitting training of the tables, so as to achieve a confusion effect. The white-box AES encryption process is divided into two stages: firstly, scrambling an addition byte replacement result of a round key for each round of AES by using random bijective mapping, so as to protect the key, and confusing combined mapping by using a neural network; and secondly, introducing a corresponding scrambling restoration operation into a column confusion operation, performing an exclusive or operation on a randomly generated key byte after performing a finite field multiplication operation, and replacing the operation with the neural network. This is also the case for the decryption process. The present invention is highly combined with a neural network, such that some steps in packet encryption are confused by using the neural network for the first time, and a reliable black-box environment is provided for some operations by means of the non-interpretability of the neural network, thereby presenting the advantages of good security, high anti-attack capability and high expansibility.
H04L 9/06 - Dispositions pour les communications secrètes ou protégéesProtocoles réseaux de sécurité l'appareil de chiffrement utilisant des registres à décalage ou des mémoires pour le codage par blocs, p. ex. système DES
Disclosed in the present disclosure are a synonym mining method and apparatus, and a storage medium and an electronic device. In the embodiments of the present disclosure, on the basis of an exposure log and a click log in a search log sequence of a user that is generated within an identical session on the basis of a search term sequence, an encoded character string of the search log sequence and a search term string of the search term sequence are determined. On the basis of the principle that "a click after term-substituted search" represents a strong probability that a search term before term substitution and a search term upon a click after the term substitution are synonyms, a search term substring is extracted from the search term string according to a preset encoded character pattern, candidate synonym pairs are mined on the basis of the search term substring, and a final synonym pair is determined on the basis of the candidate synonym pairs. In the method, a context log of a user for search terms is used in combination with search terms before and after the execution of a click behavior of the user, such that an identical search term can be avoided from having different meanings in different context, thereby improving the accuracy of synonym pair mining.
A method, a device and a medium for genome graph analysis based on in-memory computing. The method comprises the following steps: firstly, combining a linear reference genome with genetic variation to construct a genome graph; then, generating indexes for a plurality of vertices of the genome graph, and constructing an index table according to the generated indexes; then dividing the read length into a plurality of substrings with the length of k-mer, and querying the index table to obtain a seed position, generating a reference subgraph according to the seed position, and identifying a candidate mapping position according to the reference subgraph to filter a candidate mapping area; finally, using a PUM mode to run approximate string matching between the read length and all unfiltered candidate mapping positions, so as to complete the optimal alignment of a reference gene sequence and a query gene sequence.
Disclosed in the present invention is a probabilistic-diffusion-based method for protecting and recovering facial data by means of differential privacy and gradient. The method comprises: S1, acquiring a facial picture dataset, performing random sampling to obtain input data, acquiring a time tensor, performing random sampling from a standard Gaussian distribution according to the size of the input data to obtain a first Gaussian tensor, and inputting the time tensor and the first Gaussian tensor into a neural network for training to obtain an approximate Gaussian tensor and a hidden layer tensor; and S2, performing random sampling from the standard Gaussian distribution according to the size of the hidden layer tensor to obtain a second Gaussian tensor, generating an estimated picture, inputting the estimated picture into a trained neural network to obtain an approximate Gaussian tensor function, and on the basis of the second Gaussian tensor and the approximate Gaussian tensor function, performing simulated gradient descent on the estimated picture to obtain a restored facial picture.
A system for predicting disease with graph convolutional neural network based on multimodal magnetic resonance imaging, which extracts the radiomics information of multiple brain regions across modals as the features of nodes from multimodal magnetic resonance data, and extracts the connectomics information between brain regions to form an adjacency matrix. T1-weighted structural images extract cortical information through cortical reconstruction, and resting-state magnetic resonance data are used to calculate amplitude of low frequency fluctuations, regional homogeneity and functional connectivity. Through multimodal data preprocessing, image index extraction and structured data integration, multimodal unstructured magnetic resonance image data are integrated into unified graph-structured data, and the disease is predicted by a graph convolutional neural network method. The system can better integrate the cross-modal physiological indexes of multiple brain regions and the correlation between brain regions and improve prediction ability of the model and generalization ability of the model with different diseases.
The present application relates to a piezoelectric micro electric motor and a preparation method therefor. The piezoelectric micro electric motor comprises a stator and a rotor, wherein the stator comprises a controlled deformation portion, a passive deformation portion and a transmission connecting rod group. The controlled deformation portion and the passive deformation portion are provided with rotor through holes, and the rotor is arranged in at least one of the rotor through holes. The controlled deformation portion is further provided with an inverse piezoelectric portion, which is configured to undergo deformation. The transmission connecting rod group is located between the controlled deformation portion and the passive deformation portion. The transmission connecting rod group comprises at least two transmission connecting rods, and two ends of each transmission connecting rod are respectively connected to the controlled deformation portion and the passive deformation portion. The controlled deformation portion, the passive deformation portion and the transmission connecting rod group are integrally formed. According to the embodiments in the present application, machining requirements for a millimeter-sized or even micrometer-sized piezoelectric micro electric motor 10 can be met.
H02N 2/10 - Machines électriques en général utilisant l'effet piézo-électrique, l'électrostriction ou la magnétostriction produisant un mouvement rotatif, p. ex. moteurs rotatifs
36.
PANCREATIC POSTOPERATIVE DIABETES PREDICTION SYSTEM BASED ON SUPERVISED DEEP SUBSPACE LEARNING
A pancreatic postoperative diabetes prediction system based on supervised deep subspace learning. A deep convolutional neural network and the MITK software are used to obtain postoperative residual pancreas area, so as to taken as the region-of-interest. Traditional image radiomics features and deep semantic features are extracted from the residual pancreas area, and a high-dimensional image feature set is constructed. Clinical factors related to diabetes, including pancreatic excision rate, fat and muscle tissue components, demographic information and living habits are extracted, and a clinical feature set is constructed. Based on a supervised deep subspace learning network, image and clinical features are represented and fused in subspace in dimensionality reduction, while a prediction model is trained to mine sensitive features highly relevant to the prediction risk of a patient suffering postoperative diabetes mellitus with a high degree of automation and discriminative accuracy.
G16H 50/20 - TIC spécialement adaptées au diagnostic médical, à la simulation médicale ou à l’extraction de données médicalesTIC spécialement adaptées à la détection, au suivi ou à la modélisation d’épidémies ou de pandémies pour le diagnostic assisté par ordinateur, p. ex. basé sur des systèmes experts médicaux
A61B 6/00 - Appareils ou dispositifs pour le diagnostic par radiationsAppareils ou dispositifs pour le diagnostic par radiations combinés avec un équipement de thérapie par radiations
A61B 6/50 - Appareils ou dispositifs pour le diagnostic par radiationsAppareils ou dispositifs pour le diagnostic par radiations combinés avec un équipement de thérapie par radiations spécialement adaptés à des parties du corps spécifiquesAppareils ou dispositifs pour le diagnostic par radiationsAppareils ou dispositifs pour le diagnostic par radiations combinés avec un équipement de thérapie par radiations spécialement adaptés à des applications cliniques spécifiques
G06T 5/20 - Amélioration ou restauration d'image utilisant des opérateurs locaux
A method and a system for simulating magnetic resonance echo-planar imaging artifacts. Firstly, for K-space artifacts, K-space data are restored through normal magnetic resonance images, and the K-space data are modified pertinently, and then images with artifacts are reconstructed; for susceptibility artifacts, a susceptibility model is constructed through normal magnetic resonance images, and the magnetic field distribution is reconstructed, and then the images with distortion artifacts are reconstructed. According to the present disclosure, a large number of artifact data sets with different artifact types and artifact degrees can be quickly created through a small number of normal images, thus laying a foundation for the research of identifying artifacts, eliminating or weakening artifacts. A simulation algorithm is designed according to the principle of generation of EPI sequence artifacts, and the obtained images such as stripe artifacts, Moer artifacts, Nyquist artifacts, susceptibility artifacts and the like have good scientificity, accuracy and interpretability.
G01R 33/561 - Amélioration ou correction de l'image, p. ex. par des techniques de soustraction ou d'établissement de moyenne par réduction du temps de balayage, c.-à-d. systèmes d'acquisition rapide, p. ex. utilisant des séquences d'impulsions écho-planar
G01R 33/56 - Amélioration ou correction de l'image, p. ex. par des techniques de soustraction ou d'établissement de moyenne
G01R 33/565 - Correction de distorsions d'image, p. ex. dues à des inhomogénéités de champ magnétique
38.
SYSTEM FOR PRECISELY LOCATING ABNORMAL AREA OF BRAIN FIBER BUNDLE
A system for precisely locating abnormal areas of brain fiber bundles. The system extracts fiber connections of the whole brain from diffusion magnetic resonance data, and fiber bundle pathways extracts through self-defined fiber bundle pathways or based on brain fiber bundle templates. A selected fiber bundle pathway is projected on a fiber connection result of the whole brain and finely segmented. The imaging indexes such as fractional anisotropy, mean diffusivity, intra-neurite volume fraction and orientation dispersion index are calculated from diffusion magnetic resonance data, so as to obtain the imaging index of each node of each fiber bundle pathway. These imaging indexes are configured to classify the disease group and the healthy group by a machine learning method, and which nodes on which fiber bundle pathways have abnormal changes with different diseases can be precisely located.
A61B 5/055 - Détection, mesure ou enregistrement pour établir un diagnostic au moyen de courants électriques ou de champs magnétiquesMesure utilisant des micro-ondes ou des ondes radio faisant intervenir la résonance magnétique nucléaire [RMN] ou électronique [RME], p. ex. formation d'images par résonance magnétique
A61B 5/00 - Mesure servant à établir un diagnostic Identification des individus
The present invention provides a graph data set loading method and system, an electronic device, and a medium. The method comprises: acquiring a graph data set, and parsing the first N pieces of graph data to obtain preview graph data; configuring a graph data set field mapping relationship by means of an interactive mode on the basis of the preview graph data to obtain a graph data set field mapping relationship configuration file; completely parsing the graph data set again to obtain parsed graph data; reading the graph data set field mapping relationship configuration file; and reading a corresponding graph structure point/edge/attribute from the parsed graph data to obtain a graph structure in a standard format. According to the method of the present invention, graph data sets of any format can be loaded, and the loading efficiency of graph data sets during graph computation is greatly improved.
Disclosed are a model construction method and apparatus, and a storage medium and an electronic device. The model construction method comprises: acquiring a model construction request; determining a plurality of candidate architectures according to the request; for each candidate architecture, inputting the candidate architecture into a preset proxy model, and obtaining a first performance parameter of the candidate architecture in a hardware environment by means of the proxy model; determining a weight of each candidate architecture according to the respective first performance parameters of the plurality of candidate architectures, and according to the respective weights of the plurality of candidate architectures, performing screening on the plurality of candidate architectures to obtain an architecture under test; deploying a test model in the hardware environment according to the architecture under test, and acquiring a second performance parameter of the architecture under test; according to the second performance parameter of the architecture under test and the first performance parameters corresponding to the candidate architectures other than the architecture under test, performing screening on the plurality of candidate architectures to obtain a target architecture; and constructing a target model according to the target architecture.
Disclosed in the present disclosure are a search-based deep learning model deployment method and apparatus. The method comprises: acquiring a computation graph corresponding to a deep learning model; determining operators included in the computation graph, and determining a hardware resource matched with each operator; then, according to the hardware resource matched with each operator, constructing a search space; selecting a target sample from the search space, determining an operation duration corresponding to the target sample, determining a neighborhood sample corresponding to the target sample, determining an operation duration corresponding to the neighborhood sample, if the operation duration corresponding to the neighborhood sample is shorter than the operation duration of the target sample, determining the neighborhood sample as a new target sample, and continuing to determine a new neighborhood sample corresponding to the new target sample and an operation duration corresponding to the new neighborhood sample until a preset iteration termination condition is met; and, according to a finally determined allocation solution corresponding to the target sample, allocating hardware resources to the operators of the deep learning model so as to perform deployment.
A method for generating cognitive training material, a cognitive training method, an apparatus, and a medium. The method for generating cognitive training material comprises: obtaining a first feature and a second feature, the first feature comprising a multimedia material and corresponding semantic information, and the second feature comprising a magnetic resonance representation; fitting the first feature and the second feature, obtaining a semantic graph according to the fitting result and a preset brain map, and, according to the semantic graph, obtaining target semantic information corresponding to a target point; using the first feature as an input and the second feature as a constraint, training a deep learning model, and determining a weight parameter of the deep learning model; and generating a cognitive training material according to the target semantic information and the weight parameter of the deep learning model.
G16H 20/70 - TIC spécialement adaptées aux thérapies ou aux plans d’amélioration de la santé, p. ex. pour manier les prescriptions, orienter la thérapie ou surveiller l’observance par les patients concernant des thérapies mentales, p. ex. la thérapie psychologique ou le training autogène
43.
COGNITIVE TRAINING MATERIAL GENERATION METHOD, CONGNITIVE TRAINING METHOD, DIVICE, AND MEDIUM
A cognitive training material generation method, a cognitive training method, a device, and a medium are provided. The cognitive training material generation method includes: acquiring a first feature and a second feature, the first feature including a multimedia material and semantic information corresponding to the multimedia material, the second feature including a magnetic resonance representation; fitting the first feature and the second feature, obtaining a semantic map according to a fitting result and a preset brain map, and acquiring target semantic information corresponding to a target point according to the semantic map; taking the first feature as input of a deep learning model and the second feature as a constraint of the deep learning model, training the deep learning model, and determining a weight parameter of the deep learning model; generating a cognitive training material according to the target semantic information and the weight parameter of the deep learning model.
A61B 5/055 - Détection, mesure ou enregistrement pour établir un diagnostic au moyen de courants électriques ou de champs magnétiquesMesure utilisant des micro-ondes ou des ondes radio faisant intervenir la résonance magnétique nucléaire [RMN] ou électronique [RME], p. ex. formation d'images par résonance magnétique
Methods, systems, apparatus, and computer-readable media for distributed communication are provided. In one aspect, a system includes: a first Dynamic Communication Network Object (DCNO) configured on a first device and a second DCNO configured on a second device. The second DCNO is configured to, based on a notification message sent by a first worknode, allocate a target memory to store the target data in a memory of the second device, generate a read request based on the target data and the target memory, and transmit the read request to the first DCNO. The first DCNO is configured to: based on one or more properties of the target data, retrieve the target data from a memory of the first device, and write the target data to the target memory in the second device. A second worknode is configured to perform one or more data processing tasks based on the target data.
A method and a system for large-scale traffic generation based on programmable network technology, which are used for the research on network operation and maintenance and defense of attacks such as DDOS. According to the method, the required large-scale traffic is generated as required through the coordination of servers and programmable switches. The method specifically comprises the steps of designing a series of primitives which are based on intentions and are irrelevant to underlying architecture details, and reducing the description difficulty of generating large-scale traffic intentions; completing required configurations on the switch and the server by the designed cooperation mechanism of the server and programmable switch according to intentions expressed by different types of primitives, and achieving large-scale traffic generation by coordinating and utilizing server and switch resources.
H04L 69/326 - Protocoles de communication intra-couche entre entités paires ou définitions d'unité de données de protocole [PDU] dans la couche transport [couche OSI 4]
46.
MODEL TRAINING METHOD AND APPARATUS, AND MOLECULAR STRUCTURE INFORMATION RECOMMENDATION METHOD AND APPARATUS
Disclosed are a model training method and apparatus, a molecular structure information recommendation method and apparatus, a storage medium, and an electronic device. The model training method comprises: acquiring a proteolysis targeting chimera data set; according to the proteolysis targeting chimera data set, constructing three-dimensional molecular graph information of a designated proteolysis targeting chimera; inputting the three-dimensional molecular graph information of the designated proteolysis targeting chimera into a prediction model to be trained, to obtain fragment information of a molecular fragment combining with the designated proteolysis targeting chimera to have a preset medical function, and using the fragment information of the combined molecular fragment as target fragment information; training the prediction model according to deviations between multiple pieces of predicted target fragment information and fragment information of labelled molecular fragments corresponding to multiple specified proteolysis targeting chimeras.
Provided in the present invention is a method for intelligent identification of rice growth potential based on unmanned aerial vehicle (UAV) monitoring, the method including the following steps: obtaining rice plot images, labeling the images, establishing a deep convolutional neural network detection model, using the labeled rice plot images to optimize and train the model, inputting the rice plot images to be measured into the trained model, and detecting a location of a rice plot target frame in each image; selecting a target frame with the largest area in each rice plot image, and pre-processing a rice plot image in the target frame; and calculating a vegetation coverage rate of the pre-processed rice plot image, and determining a level of rice growth potential according to the vegetation coverage rate. Also provided in the present invention is a system for intelligent identification of rice growth potential based on unmanned aerial vehicle (UAV) monitoring. The method of the present invention has the advantages of simplicity, high precision, fast speed and low cost in the identification of the rice growth potential, and can be widely used in automatic and intelligent production management of agriculture.
G06K 9/00 - Méthodes ou dispositions pour la lecture ou la reconnaissance de caractères imprimés ou écrits ou pour la reconnaissance de formes, p.ex. d'empreintes digitales
G06V 10/26 - Segmentation de formes dans le champ d’imageDécoupage ou fusion d’éléments d’image visant à établir la région de motif, p. ex. techniques de regroupementDétection d’occlusion
G06V 10/28 - Quantification de l’image, p. ex. seuillage par histogramme visant à discriminer entre les formes d’arrière-plan et d’avant-plan
G06V 10/74 - Appariement de motifs d’image ou de vidéoMesures de proximité dans les espaces de caractéristiques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
G06V 20/17 - Scènes terrestres transmises par des avions ou des drones
G06V 20/70 - Étiquetage du contenu de scène, p. ex. en tirant des représentations syntaxiques ou sémantiques
B64U 101/00 - Véhicules aériens sans pilote spécialement adaptés à des utilisations ou à des applications spécifiques
B64U 101/30 - Véhicules aériens sans pilote spécialement adaptés à des utilisations ou à des applications spécifiques à l’imagerie, à la photographie ou à la vidéographie
48.
NEURAL MODEL STORAGE SYSTEM AND METHOD FOR OPERATING SYSTEM OF BRAIN-INSPIRED COMPUTER
A neural model storage system and method for an operating system of a brain-inspired computer are provided. The method includes: storing a neural model on three computing nodes, selecting the computing nodes by dynamically calculating a weight according to the number of idle cores of the first computing node, the number of failures of the first computing node, and failure time of the first computing node in each failure thereof, reading the neural model in the same computing node or cross-computing node, recovering from failures of non-master nodes, recovering from failures of the master node, and recovering from a whole machine restart or failure.
An integrated pressure/temperature/proximity multi-parameter sensor based on photoelectric fusion. The sensor comprises a flexible optical waveguide based on an optical mechanism and a flexible interdigital electrode thin film based on an electrical mechanism, which thin film is wound around the flexible optical waveguide, wherein the flexible optical waveguide consists of two optical fibers which are inserted into a silicone tube and are spaced apart from each other. The sensor implements crosstalk-free self-decoupling sensing of three parameters, i.e., pressure, temperature and proximity, by means of a multi-dimensional response signal based on photoelectric fusion, wherein pressure is measured in the form of light intensity by means of an optical waveguide loss, temperature is measured in the form of resistance by means of a thermal resistance effect of an electrode, and object proximity is measured in the form of capacitance by means of an edge electric field of an interdigital electrode. Therefore, pressure, temperature and proximity can be simultaneously monitored without causing signal crosstalk, the structure is compact, the preparation is simple, and a complex system integration and decoupling algorithm is not required. Further provided is a preparation method for an integrated multi-parameter sensor.
G01D 21/02 - Mesure de plusieurs variables par des moyens non couverts par une seule autre sous-classe
G01L 1/24 - Mesure des forces ou des contraintes, en général en mesurant les variations des propriétés optiques du matériau quand il est soumis à une contrainte, p. ex. par l'analyse des contraintes par photo-élasticité
G01K 7/18 - Mesure de la température basée sur l'utilisation d'éléments électriques ou magnétiques directement sensibles à la chaleur utilisant des éléments résistifs l'élément étant une résistance linéaire, p. ex. un thermomètre à résistance de platine
G01D 5/24 - Moyens mécaniques pour le transfert de la grandeur de sortie d'un organe sensibleMoyens pour convertir la grandeur de sortie d'un organe sensible en une autre variable, lorsque la forme ou la nature de l'organe sensible n'imposent pas un moyen de conversion déterminéTransducteurs non spécialement adaptés à une variable particulière utilisant des moyens électriques ou magnétiques influençant la valeur d'un courant ou d'une tension en faisant varier la capacité
50.
BASIC RADIATING UNIT OF ANTENNA FOR MILLIMETER WAVE OR ABOVE FREQUENCY BAND
The present disclosure relates to communications technologies, and provides a single polarized ME dipole and a basic radiator of a phased array antenna for millimeter wave or above frequency band. The radiator is an "array of the ME dipoles" with higher dielectric constant, and can be manufactured with the multilayer PCB or LTCC process. The ME dipole comprises an electric dipole made by two parts of a first metal sheet located on a top substrate layer, a magnetic dipole made by first vias connected to the first metal sheet and a third metal sheet located on a bottom substrate layer, a balun made by an other part of the first metal sheet and a third via, and a reflecting part made by a part of the third metal sheet. The substrate layers are stacked in a multilayer structure to facilitate arranging the metal sheets and the vias.
Disclosed in the present invention are a silk fibroin-based multi-responsive soft actuator and a manufacturing method therefor. The soft actuator comprises a silk fibroin membrane and a flexible substrate, the silk fibroin membrane being arranged on and tightly bonded with the flexible substrate to form a double-layer membrane structure, thermal expansion coefficients of the silk fibroin membrane and the flexible substrate being different. The manufacturing method for a soft actuator comprises: performing plasma processing on a flexible substrate; then scrap-coating the flexible substrate with a silk fibroin wet membrane, drying same to obtain a silk fibroin membrane, the silk fibroin membrane together with the flexible substrate forming a double-layer membrane; soaking the double-layer membrane into water, and then drying same; and integrally or locally soaking in a calcium chloride aqueous solution the silk fibroin membrane in the dried double-layer membrane, then taking out same, and drying same to obtain a soft actuator. Deformation of the soft actuator of the present invention can be driven by tiny energy inputs, making energy consumption low. The soft actuator exhibits better reversibility, has a larger deformation angle, achieves reprogrammable deformation in different regions, and is responsive to multiple stimuli.
F03G 7/06 - Mécanismes produisant une puissance mécanique, non prévus ailleurs ou utilisant une source d'énergie non prévue ailleurs utilisant la dilatation ou la contraction des corps produites par le chauffage, le refroidissement, l'humidification, le séchage ou par des phénomènes similaires
52.
PAPER CLASSIFICATION METHOD AND APPARATUS, STORAGE MEDIUM, AND ELECTRONIC DEVICE
The present disclosure discloses a paper classification method and apparatus, a storage medium, and an electronic device. According to embodiments of the present disclosure, by means of a paper category prediction model, for each node in a topological graph comprising association information of a paper, a preset number of feature adjustments are performed on an initial node feature of the node on the basis of a node feature of the node, a node feature of a neighbor node and an attention weight between the node and the neighbor node, to obtain a final node feature of the node, and the category of the paper is predicted on the basis of the final node feature of a paper node. The association information comprises the text of the paper, the text of a cited paper, the author of the paper, the institution of the author and the like.
A robot control method and apparatus, and a storage medium and an electronic device. The robot control method comprises: when a biped robot moves by means of mechanical feet, determining a mechanical-foot support state corresponding to each time stage (S101); according to the mechanical-foot support state corresponding to each time stage, determining a movement trajectory of a zero moment point (ZMP) between the mechanical feet, which ZMP corresponds to the biped robot (S102); for each mechanical foot of the biped robot, determining first pose information corresponding to the mechanical foot in a first time stage, wherein the first time stage comprises a plurality of time stages in which the mechanical foot is in a support state, and according to the first pose information, determining a motion trajectory corresponding to the mechanical foot (S103); according to the movement trajectory of the ZMP and the motion trajectory corresponding to each mechanical foot, determining a motion trajectory of the centroid of the biped robot (S104); and on the basis of the motion trajectory of the centroid and the motion trajectory corresponding to each mechanical foot, determining a joint angle trajectory corresponding to each target joint of the biped robot, and according to the joint angle trajectory, determining motion planning data for controlling the biped robot, so as to control, according to the motion planning data, the biped robot to move by means of the mechanical feet (S105).
G05B 19/042 - Commande à programme autre que la commande numérique, c.-à-d. dans des automatismes à séquence ou dans des automates à logique utilisant des processeurs numériques
54.
DEVICE AND METHOD OF INTRA-CHIP ROUTING OF NEURAL TASKS FOR OPERATING SYSTEM OF BRAIN-INSPIRED COMPUTER
A method and a device of intra-chip routing of neural tasks for an operating system of a brain-inspired computer are provided. The method includes determining an area defined by target cores, and determining target cores in a row furthest from an edge routing area; determining whether the target cores need to be configured with relay routing; searching nearest edge routing cores in the edge routing area for all the target cores in the area defined by the target cores; configuring the target cores in a far-to-near principle; and searching relay routing cores and the nearest edge routing cores by a shortest path manner and a maximum step length of a single routing manner.
H04L 45/122 - Évaluation de la route la plus courte en minimisant les distances, p. ex. en sélectionnant une route avec un nombre minimal de sauts
H04L 49/109 - Éléments de commutation de paquets caractérisés par la construction de la matrice de commutation intégrés sur micropuce, p. ex. interrupteurs sur puce
55.
LONG-ENDURANCE OPERATION HUMAN–COMPUTER INTERACTION METHOD AND APPARATUS BASED ON DYNAMIC TIME PRESSURE
The present invention belongs to the technical field of human–computer ergonomics. Disclosed are a long-endurance operation human–computer interaction method and apparatus based on a dynamic time pressure. The method specifically comprises: constructing a highest performance-time pressure-time function curve according to a performance-time pressure function curve corresponding to a long-endurance operation; after the current long-endurance operation ends, calculating a response time of the next long-endurance operation according to a historical response time of human–computer interaction; according to the highest performance-time pressure-time function curve, calculating a highest performance value corresponding to the starting time of the next long-endurance operation and a time pressure corresponding to the highest performance value; and according to the response time of the next long-endurance operation and the time pressure, calculating an optimal reserved time corresponding to the highest performance value. The method and apparatus can improve the performance level of long-endurance operation human–computer interaction, reduce human errors caused by accumulated fatigue and excessively short reserved operation time, and improve flight safety.
A current control method and device, an electronic device, and a storage medium. The current control method comprises: acquiring the current first temperature of a motor and the current second temperature of a driver, wherein the driver is used for driving the motor to operate; determining a predicted operation parameter of the motor on the basis of the first temperature and the second temperature, wherein the predicted operation parameter is used for representing an operation parameter when the motor and the driver operate from the current temperatures to a preset temperature; determining a predicted filtering current of a filter on the basis of the predicted operation parameter, wherein the filter is used for filtering an output current of the driver; and controlling the current of the motor on the basis of the predicted filtering current.
A method and apparatus for predicting a parking-space idle rate, and a medium and a device. The method comprises: acquiring parking-space idle rates, at a plurality of moments before a moment to be subjected to prediction, of parking lots in a region to be subjected to prediction, and using the parking-space idle rates as historical idle rates of the parking lots (S102); inputting the historical idle rates of the parking lots into a feature extraction network, so as to obtain a first feature, wherein the first feature is used for representing the relationship between the historical idle rates of the parking lots and time (S104); inputting a spatial relationship graph and the first feature into a graph fusion network, so as to obtain a fused feature (S108); and then inputting the fused feature into a result prediction network, so as to obtain parking-space idle rates of the parking lots in said region at the moment to be subjected to prediction (S110). On the basis of the relationship between parking-space idle rates and time, the relationship between the parking-space idle rates and space, and a potential interaction relationship between the parking-space idle rates and time and space, prediction results of parking-space idle rates of a plurality of parking lots in a region to be subjected to prediction are obtained at the same time, thereby improving the prediction efficiency of the parking-space idle rates and the accuracy of the prediction results.
A method, a system, a device, and a storage medium for operation resource placement of deep learning are provided. The method includes: acquiring training operations to be placed and corresponding priorities; based on an order of the priorities, selecting a network structure for operation placement according to required resource amount of the training operations in sequence; the network structure including a server, a top of rack, a container group set denoted as Podset and a trunk layer switch; based on the selected network structure, taking a transmission amount of network data in a training process as an optimization target to perform minimization optimization, and obtaining a corresponding operation placement scheme.
A deep learning operation resource placement method, a system, a device, and a storage medium. The method comprises: acquiring a training operation to be placed and a corresponding priority level; on the basis of a priority level order, sequentially selecting a network structure for operation placement according to a resource demand the training operation, the network structure comprising a server, a top switch, a container group set (Podset) and a backbone layer switch; on the basis of the selected network structure, performing minimization optimization using a network data transmission amount in a training process as an optimization target, to obtain a corresponding operation placement solution.
Disclosed in the present invention are a breeding cross-generation phenotype prediction method and system based on ensemble learning, and an electronic device. The method comprises: acquiring genotype data of a high-generation crop and a corresponding post-generation crop, and collecting target phenotype data of the high-generation crop; calculating an evaluation function on the basis of a genetic algorithm, and according to the evaluation function, screening the genotype data to select a genotype data subset, which has a genetic correlation with the corresponding post-generation crop, in the high-generation crop; training several different machine learning models by means of the subset; calculating evaluation indexes of the machine learning models, sorting the machine learning models, and selecting the first K machine learning models to be basic learners; stacking the K basic learners on the basis of an ensemble learning method, and training same, so as to obtain a meta-learner; and inputting the genotype data of the post-generation crop into the basic learners to obtain metadata, and then inputting the metadata into the meta-learner to obtain predicted target phenotype data of the post-generation crop.
G16B 25/10 - Profilage de l’expression de gènes ou de protéinesEstimation ou normalisation de ratio d’expression
G16B 40/00 - TIC spécialement adaptées aux biostatistiquesTIC spécialement adaptées à l’apprentissage automatique ou à l’exploration de données liées à la bio-informatique, p. ex. extraction de connaissances ou détection de motifs
G06N 20/20 - Techniques d’ensemble en apprentissage automatique
61.
ELECTRICAL CONTROL SYSTEM, METHOD AND APPARATUS FOR HUMANOID ROBOT
An electrical control system for a humanoid robot. The humanoid robot comprises a robot body and a drive mechanism for driving the robot to move. The electrical control system comprises a positioning fusion module, an understanding decision-making module, a motion control module and a joint drive module. Efficient multi-task data processing between the modules is implemented by means of a multi-core heterogeneous processing mode and a private 5G network, high-performance computing and analysis capabilities are implemented in the case of a low power consumption, and high-reliability execution of non-real-time scenario understanding and real-time motion control tasks is implemented, such that it is ensured that the robot body implements multi-scenario efficient operations with a low computing power and a low power consumption capability, thereby reducing the costs and the system complexity. The present invention further relates to an electrical control method and apparatus for a humanoid robot.
A prediction method and apparatus, a readable storage medium and an electronic device. The method comprises: on the basis of the molecular structure of a molecule to be predicted, by means of a graph neural network model, determining specified sub-graphs corresponding to sub-structures of said molecule, so as to determine specified properties of the specified sub-graphs on the basis of specified features corresponding to the specified sub-graphs and preset characterization features of each specified property, thus predicting molecular properties of said molecule. Therefore, said molecule has corresponding molecular properties since said molecule comprises sub-structures having specified properties. Obviously, the prediction method provides explainability for said molecule to have the corresponding molecular properties, thus ensuring the credibility of the prediction result.
A method and system for obtaining link node data is provided. The method is applied to an industrial control ring network system including a master station device and a plurality of node devices, and includes: after receiving a link fault message, the master station device determines to transmit a first link location acquisition message to a first network port and/or a second network port based on respective on-off states; each node device processes the first link location acquisition message to obtain and transmit a second link location acquisition message; the master station device parses the second link location acquisition message and transmits a first node data acquisition message; the node device processes the first node data acquisition message to obtain and transmit a second node data acquisition message; and the master station device parses the second node data acquisition message to obtain data of at least one node device.
H04L 41/0668 - Gestion des fautes, des événements, des alarmes ou des notifications en utilisant la reprise sur incident de réseau par sélection dynamique des éléments du réseau de récupération, p. ex. le remplacement par l’élément le plus approprié après une défaillance
64.
HETEROGENEOUS NETWORK-BASED AUTONOMOUS VEHICLE REMOTE CONTROL DEVICE AND METHOD
Disclosed are a heterogeneous network-based autonomous vehicle remote control device and method. The device comprises a vehicle information acquisition module, a first message sending module, a first message receiving module, and a first remote control module. According to the present invention, a remote control failure is avoided or the possibility of a remote control failure is greatly reduced by not taking an area not supporting remote control into consideration during vehicle path planning. Additionally, heterogeneous network resources are rationally utilized on a vehicle travel path, so that real-time performance of a remote control end in acquiring vehicle-related information is improved, thereby effectively improving the availability and reliability of remote control and the safety of vehicle driving.
G05B 19/042 - Commande à programme autre que la commande numérique, c.-à-d. dans des automatismes à séquence ou dans des automates à logique utilisant des processeurs numériques
B60W 60/00 - Systèmes d’aide à la conduite spécialement adaptés aux véhicules routiers autonomes
H04W 4/44 - Services spécialement adaptés à des environnements, à des situations ou à des fins spécifiques pour les véhicules, p. ex. communication véhicule-piétons pour la communication entre véhicules et infrastructures, p. ex. véhicule à nuage ou véhicule à domicile
An earthquake prediction method, comprising: before a seismic wave propagates to a fortified region, detecting a signal of the seismic wave by means of a seismic station, predicting a focus location and a magnitude of the seismic wave on the basis of the detected signal of the seismic wave and station information of the seismic station, and then sending prompt information on the basis of the predicted focus location and the predicted magnitude. The present invention further provides an earthquake prediction apparatus, a computer readable storage medium, and an electronic device.
A method for suppressing the residual vibration of a robotic arm, comprising: acquiring dynamic features of a robotic arm under a preset pose and a preset load, and establishing a corresponding data set among the dynamic features, the preset pose, and the preset load; training a deep neural network model on the basis of the data set; according to a target pose and a target load of the robotic arm, predicting a target dynamic feature of the robotic arm by using the trained deep neural network model; and according to the target dynamic feature, designing a vibration suppressor to work in conjunction with a motion controller to control the robotic arm to move to the target pose and to suppress residual vibration. According to the method, the online real-time prediction of the dynamic features can be achieved at low calculation cost, the vibration suppressor is designed adaptively, and the vibration suppression in an open working scenario is achieved. Further provided are a control device, a storage medium, a robotic arm assembly, and a computer-readable storage medium.
The present application discloses a method and a system for analyzing and predicting a vehicle stay behavior based on multi-task learning, and the method includes the following steps: acquiring vehicle GPS and OBD data including a vehicle ID, a travel start time, a start longitude, a start latitude, an end time, an end longitude, and an end latitude after desensitization; preprocessing vehicle GPS and OBD data to obtain vehicle stay behavior data including stay location and stay duration; extract a spatial-temporal characteristic of the preprocessed vehicle stay behavior data by a deep recurrent neural network; inputting the spatial-temporal characteristic into a multi-task learning and predicting network, and obtaining the correlation between a stay location prediction task and the stay duration prediction task based on the historical stay behavior of the vehicle through the multi-task learning and predicting network to predict the stay location and stay duration.
The present disclosure discloses an autonomous vehicle remote control apparatus and a method based on heterogeneous networks. The apparatus comprises a vehicle information acquisition module, a first message sending module, a first message receiving module and a first remote control module. According to the present disclosure, the possibility of failure of remote control is avoided or greatly reduced by bypassing the area where the network quality does not support remote control when planning a vehicle path, heterogeneous network resources are reasonably utilized on the vehicle driving path, the real-time performance of obtaining vehicle-related information by a remote control terminal is improved, and the availability and reliability of remote control and the safety of vehicle driving are effectively enhanced.
G05D 1/00 - Commande de la position, du cap, de l'altitude ou de l'attitude des véhicules terrestres, aquatiques, aériens ou spatiaux, p. ex. utilisant des pilotes automatiques
B60W 60/00 - Systèmes d’aide à la conduite spécialement adaptés aux véhicules routiers autonomes
H04L 67/12 - Protocoles spécialement adaptés aux environnements propriétaires ou de mise en réseau pour un usage spécial, p. ex. les réseaux médicaux, les réseaux de capteurs, les réseaux dans les véhicules ou les réseaux de mesure à distance
69.
METHOD AND APPARATUS FOR IDENTIFYING NON-COMPLIANT PRODUCTS, COMPUTER DEVICE, AND STORAGE MEDIUM
The present invention provides a method and apparatus for identifying non-compliant products, a computer device, and a storage medium. The method comprises: constructing a multi-modal knowledge graph according to a multi-modal knowledge graph dataset, and extracting visual features of visual modal entities and text features of text modal entities in the knowledge graph; acquiring a product image and a product text according to a database; generating a product visual feature according to the product image; generating a product text feature according to the product text; according to the visual features and text features and the product visual features and product text features, linking the product image and product text to the knowledge graph by using an entity linking method; and according to the linked knowledge graph, acquiring the correlation between the product image and product text so as to determine the compliance of the product. By introducing a knowledge graph, the present invention achieves reasonable inference of product non-compliance risks on the basis of the knowledge, and features such significant advantages as broad coverage, cost efficiency, high robustness, and the like.
The present disclosure provides a data processing system and method. The system comprises a client interaction module, a subscription publishing module, a storage module, and a sub-database management module. The client interaction module can receive an interaction request sent by a client, and analyze the interaction request to obtain an analysis result; and determine, according to the analysis result, a process type needing to be started so as to start a response process of the process type, and re-package and send the interaction request to the response process, wherein the process type comprises a first process type, a second process type and a third process type respectively corresponding to the subscription publishing module, the storage module and the sub-database management module.
A system on wafer assembly structure and an assembly method thereof. The system on wafer assembly structure comprises: a wafer layer, a dielectric layer and a circuit board layer sequentially stacked, and each provided with a bonding region, a testing region and an alignment region, respectively, a first assembly, and a second assembly, wherein the first assembly is arranged on one side of the wafer layer far away from the dielectric layer, and comprises a bearing portion and at least one latch portion connected with each other, and the bearing portion is detachably connected with the wafer layer. The second assembly is at least partially arranged around the first assembly. The second assembly has a hole portion for accommodating a latch portion, and the inner diameter of the hole portion is larger than the outer diameter of the latch portion.
A multi-user flexible Ethernet fine granularity time slot allocation method and apparatus. The method is specifically: providing two time slot resource allocation and deployment schemes according to user demands, where the first scheme performs time slot allocation based only on an objective of global minimization of jitter of time slot allocated to all the users and can improve resource allocation equity and improve whole performance of a network. The second scheme performs weighted sum of delay and jitter minimization based time slot allocation on each user according to the quantity of time slots required for the users in sequence from large to small on the premise of most user input data in the current time slot assignment period being transmitted in the current time slot assignment period.
The present application discloses a task execution method and apparatus, a storage medium, and an electronic device. The task execution method comprises: determining a computing task corresponding to each network layer in a target model, and determining device information corresponding to a plurality of computing devices participating in target model computing; for each network layer, according to the times of computing involved when executing the computing task corresponding to the network layer and the device information of the plurality of computing devices, determining a computing duration required when the plurality of computing devices execute the computing task corresponding to the network layer; according to at least one of the computing duration, a duration of data transmission between a computing device corresponding to the previous network layer and another computing device, the memory space required by data of the network layer, and the remaining memory of each computing device, determining a target device corresponding to the network layer; and executing the computing task by means of the target device after receiving an execution request of the computing task corresponding to each network layer.
The present invention discloses the low-stress niobium nitride (NbN) superconducting thin film and preparation method and application thereof. The preparation method includes the following steps: providing the metal Nb target and the Si-based substrates, fixing the Si-based substrate at room temperature, adjusting the mass flow ratio of N2/Ar to 20%-50%, the sputtering power to 50-400 W and the deposition pressure to 3.0-10.0 mTorr, NbN superconducting thin films with a stress range of-500 MPa˜500 MPa and a thickness of 70-150 nm were deposited on Si-based substrates. By synergistically controlling the mass flow rate ratio of N2/Ar, sputtering power, and deposition pressure, low stress NbN superconducting thin films can be easily and efficiently prepared. The stress range of the prepared NbN superconducting thin films meets the preparation requirements of superconducting dynamic inductance detectors, and can be mass-produced.
C23C 14/06 - Revêtement par évaporation sous vide, pulvérisation cathodique ou implantation d'ions du matériau composant le revêtement caractérisé par le matériau de revêtement
C23C 14/35 - Pulvérisation cathodique par application d'un champ magnétique, p. ex. pulvérisation au moyen d'un magnétron
G01J 5/20 - Pyrométrie des radiations, p. ex. thermométrie infrarouge ou optique en utilisant des détecteurs électriques de radiations en utilisant des éléments résistants, thermorésistants ou semi-conducteurs sensibles aux radiations, p. ex. des dispositifs photoconducteurs
75.
METHOD, DEVICE, COMPUTER EQUIPMENT AND STORAGE MEDIUM FOR IDENTIFYING ILLEGAL COMMODITY
A method, a device, computer equipment and a storage medium for identify an illegal commodity. The method comprises: firstly, constructing a multi-modal knowledge graph according to a multi-modal knowledge graph data set, and extracting visual features of all visual modality entities and text features of all text modality entities in the knowledge graph; then obtaining a commodity image and a commodity text according to a database; then, generating commodity visual feature according to the commodity image; then generating the commodity text feature according to the commodity text; secondly, according to the visual features and text features, as well as the commodity visual feature and the commodity text feature, linking the commodity image and the commodity text to the knowledge graph by using an entity linking method; finally, obtaining the correlation between the commodity image and the commodity text according to the linked knowledge graph to determine the illegality of the commodity.
HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY (Chine)
ZHEJIANG LAB (Chine)
Inventeur(s)
Zheng, Long
He, Haiheng
Liao, Xiaofei
Jin, Hai
Chen, Dan
Huang, Yu
Abrégé
A method for incremental metapath storage and dynamic maintenance is provided, which includes, reformatting metapath instances, from a designated heterogeneous graph and of a designated metapath type, into path graphs; executing graph updating tasks and performing dynamic maintenance on the updated path graphs, traversing the path graph to obtain the location of metapath updates and update the path graph; for metapaths with length greater than 2 and with symmetrical central portion, central merge operation is performed to simplify path graph and perform subsequent restoration operation; and directly perform restoration operation on path graphs that do not meet the merging conditions. The present disclosure utilizes characteristics of graph update to obtain locality of metapath updates, and combines internal relationship characteristics of metapath instances to greatly speed up metapath generation and achieve real-time inference of dynamic heterogeneous graph models.
A macro-micro collaborative topology design method for a thin-walled structure, comprising the following steps: (1) setting a working condition I and a working condition II; (2) combining the working condition I and the working condition II, designing an initial shank model, and carrying out finite element calculation; (3) reconstructing a lightweight shank model by using a topology design method; (4) for the reconstructed lightweight shank model, keeping the thickness of the shell on the outer surface to be 2 mm, and configuring the core as a lattice structure; (5) performing variable thickness design on the shell according to the finite element calculation result; (6) performing variable cell side length or rod diameter change design on the lattice structure of the core; and (7) performing dynamic iteration on the topology design result to obtain a uniform stress field. Also provided is a robot shank model using the macro-micro collaborative topology design method for a thin-walled structure. According to the method and the model, on the basis of realization of integration of structural parts and high attractiveness, the macro-micro collaborative topology design reduces the weight by 30% in total, and moreover, the strength and the rigidity meet the use requirements of a robot.
G06F 30/23 - Optimisation, vérification ou simulation de l’objet conçu utilisant les méthodes des éléments finis [MEF] ou les méthodes à différences finies [MDF]
78.
MULTI-CHIP COMMUNICATION METHOD BASED ON FPGA MASTER CONTROLLER AND NEURAL NETWORK ALGORITHM
Disclosed in the present invention is a multi-chip communication method based on an FPGA master controller and a neural network algorithm. Image processing based on a neural network algorithm is completed by means of design of an original data frame, a status frame, a hierarchical data frame, a hierarchical weight frame, a calculation result frame, a hierarchical data request frame, a hierarchical weight request frame, a calculation result request frame, and an operation status request frame, and by means of scheduling of sending and receiving processes. According to the present invention, neural network algorithm-based communication of multi-level data structures and multiple data types is ensured, sending and receiving of data required by the master controller and chips in a multi-chip system are accurately scheduled, and data request commands are sent; the present invention plays a very active role in giving feedback on receiving, sending, chip running statuses, errors that occur, and error types.
Disclosed in the present invention are an efficient serverless resource allocation method and system based on reinforcement learning. In the method, by means of the observation of a relationship between a tail latency, a decision frequency and a resource efficiency, a set performance delay objective is guaranteed while the resource allocation consumption of a serverless system is minimized. The method makes full use of the advantage of efficient resource management brought about by high-frequency management, and by means of observing the state of each request, a decision is made, by using a reinforcement learning model, on a resource configuration of an instance for processing the request. By means of taking into account the characteristic of multi-stage operation of a function workflow and a lightweight design for a decision-making model, a high-frequency control layer conceals time overheads and reduces resource overheads. Compared with the latest workflow task scheduling system, the present invention improves the CPU utilization rate, and provides a delay service level objective (SLO) guarantee for 99% requests, thereby reducing an end-to-end delay variance.
Huazhong University of Science and Technology (Chine)
Inventeur(s)
Zhang, Zhan
Zhang, Yu
Zhao, Jin
Wu, Haifei
Abrégé
The present disclosure provides a data storage system, including data cache module, data processing module, and a persistent memory. The data cache module includes an on-chip mapping data cache and an on-chip counter cache, where the mapping data cache is configured to cache mapping data, and when the free space of the mapping data cache is less than a preset threshold, the least recently used mapping data cache line will be evicted from the cache and written back to the persistent memory. The data processing module encrypts/decrypts persistent memory data by using their counters, and accesses the persistent memory blocks indicated by their corresponding mapping data. The persistent memory comprises the first and second storage regions for the latest checkpoint data and modified working data in the current checkpoint interval respectively.
G06F 12/0804 - Adressage d’un niveau de mémoire dans lequel l’accès aux données ou aux blocs de données désirés nécessite des moyens d’adressage associatif, p. ex. mémoires cache avec mise à jour de la mémoire principale
81.
TEXT FEATURE EXTRACTION METHOD AND SYSTEM, AND ELECTRONIC DEVICE AND MEDIUM
ininininininin being positive integers; and constructing a text feature extraction network, and extracting a tensor sequence feature by means of the text feature extraction network. In the method of the present invention, each element feature and a text feature in a text sequence are expanded in two dimensions, such that a feature tensor can include more semantic information than a word vector, such as an orientation and a shape; moreover, a tensor sequence is processed in combination with a text feature extraction network having local connection characteristics, thereby enhancing the interpretability and semantic capacity of the features.
A robotic extension and retraction apparatus capable of rapid movement, comprising a fixed frame. The fixed frame comprises a mounting bottom plate (11), a top plate (13) and two mounting side plates (12) on two sides; a head mounting member (32) used for mounting a head is provided on the inner side of the top plate (13); the top plate (13) is provided with a head avoiding hole (15) for the head to pass through; a limb mounting member (43) used for mounting a limb is provided on the inner side of each mounting side plate (12); each mounting side plate (12) is provided with a limb avoiding hole (14) for the limb to pass through; a driving apparatus used for driving the head mounting member (32) and the limb mounting member (43) to move so as to enable the head and the limb to be at least partially retracted into the fixed frame is provided in the fixed frame; and the driving apparatus comprises a steering engine (21) fixed on the mounting bottom plate (11), and rotating supports (22) driven by the steering engine (21) to rotate so as to drive the head mounting member (32) and the limb mounting members (43) to move. Further provided is a robot capable of rapid movement.
B25J 19/00 - Accessoires adaptés aux manipulateurs, p. ex. pour contrôler, pour observerDispositifs de sécurité combinés avec les manipulateurs ou spécialement conçus pour être utilisés en association avec ces manipulateurs
83.
RADIO ASTRONOMY INTERFERENCE SIGNAL REDUCTION METHOD, APPARATUS, AND SYSTEM AND COMPUTER DEVICE
The present disclosure relates to a radio astronomy interference signal reduction method, apparatus, and system and a computer device. The method includes: acquiring at least two beam signals, and obtaining a covariance matrix between all the beam signals according to the beam signals, wherein the beam signals include subsignals from at least two directions; performing eigen-decomposition processing on the covariance matrix to obtain signal eigenvectors and eigenvalues corresponding to the subsignals, and extracting interference eigenvectors corresponding to interference signals from the signal eigenvectors based on the eigenvalues; and obtaining a target reduction result for the interference signals according to the signal eigenvectors and the interference eigenvectors. With the method, the problem of low accuracy of radio astronomy interference signal reduction in the related art can be solved.
H04B 7/08 - Systèmes de diversitéSystèmes à plusieurs antennes, c.-à-d. émission ou réception utilisant plusieurs antennes utilisant plusieurs antennes indépendantes espacées à la station de réception
H04B 17/21 - SurveillanceTests de récepteurs pour l’étalonnageSurveillanceTests de récepteurs pour la correction des mesures
84.
DATA PROCESSING METHOD AND APPARATUS, STORAGE MEDIUM AND ELECTRONIC DEVICE
The present disclosure relates to a data processing method and apparatus, a storage medium and an electronic device. In the method, after a switch chip receives a data frame, the data frame is analyzed by a data analysis model deployed in a data processing unit and based on an analysis result, a processing policy for the data frame is determined, and the switch chip processes the data frame based on the processing policy.
Disclosed in the present disclosure are a data processing method and apparatus, and a storage medium and an electronic device. The data processing method comprises: acquiring each piece of data to be processed; determining whether a data processing model can process a set number of pieces of data to be processed under the current processing process, and if the data processing model cannot process the set number of pieces of data to be processed under the current processing process, retrieving data processing times of the data processing model under different configuration combinations; for the data processing time under each configuration combination, determining, as a target data volume, a data volume that the data processing model can process within the data processing time; taking the data processing model being capable of processing the set number of pieces of data to be processed as a target, selecting a target configuration combination from among the configuration combinations according to the target data volume; and creating a processing process under the target configuration combination for the data processing of the data to be processed.
The present disclosure provides a data processing system and a data processing method. The system includes: a client interaction module, a subscribing and publishing module, a storage module, and a sub-database management module. The client interaction module is configured to: receive an interaction request sent by a client, analyze the interaction request to obtain an analyzing result, and based on the analyzing result, determine a process type to be started and start a response process of the process type, and repackage the interaction request and send the repackaged interaction request to the response process, where the process type includes a first process type corresponding to the subscribing and publishing module, a second process type corresponding to the storage module and a third process type corresponding to the sub-database management module.
H04L 67/1097 - Protocoles dans lesquels une application est distribuée parmi les nœuds du réseau pour le stockage distribué de données dans des réseaux, p. ex. dispositions de transport pour le système de fichiers réseau [NFS], réseaux de stockage [SAN] ou stockage en réseau [NAS]
A vehicle stopping behavior analysis and prediction method and system based on multi-task learning. The method comprises the following steps: collecting vehicle GPS and OBD data, which comprises a vehicle ID, a travel start time, a start longitude, a start latitude, an end time, an end longitude and an end latitude after vehicle desensitization; preprocessing the vehicle GPS and OBD data, so as to obtain vehicle stopping behavior data, which comprises a stopping place and a stopping duration, and by using a deep recurrent neural network, performing spatio-temporal feature extraction on a vehicle stopping behavior after preprocessing is performed; and inputting a spatio-temporal feature into a multi-task learning and prediction network, and on the basis of a historical stopping behavior of a vehicle, acquiring the correspondence between a stopping place prediction task and a stopping duration prediction task by means of the multi-task learning and prediction network, so as to predict the stopping place and the stopping duration. The method can provide support for application scenarios such as point-of-interest recommendation, personalized travel formulation, targeted advertisement delivery and smart parking.
G06Q 10/04 - Prévision ou optimisation spécialement adaptées à des fins administratives ou de gestion, p. ex. programmation linéaire ou "problème d’optimisation des stocks"
Huazhong University of Science and Technology (Chine)
Inventeur(s)
Zhang, Yu
Qi, Hao
Luo, Kang
Zhao, Jin
Zhang, Zhan
Abrégé
Systems, methods, devices and storage media for graph data processing are provided. In one aspect, a graph data processing system includes a memory and a plurality of processing units, and each processing unit is provided with a decision module. Each processing unit is configured to determine set operations required for extracting one or more subgraphs matching a specified graph pattern from target graph data according to a preset graph pattern matching algorithm. Then, for each set operation, the decision module is configured to determine a cost value corresponding to a performance of the processing unit occupied to execute the set operation in accordance with different execution policies, and further select a target execution policy with a smallest cost value to execute the set operation.
A content caching method for a satellite terrestrial integrated network with differentiated interests and access modes, including: collecting user proportion information, content popularity distribution information, cache capacity information, and link delay information in the satellite terrestrial integrated network; determining a content caching problem of the collaboration of base stations and a satellite for a delay optimization based on differentiated user interests and differentiated access modes; solving the content caching problem to obtain a content caching strategy of the collaboration of the base stations and the satellite; and based on the obtained content caching strategy, caching contents in the base stations and the satellite to provide a content service for base station access users and satellite access users in the satellite terrestrial integrated network.
The present invention provides a satellite-ground cooperative network content caching method in a differentiated interest and access mode. The method comprises: collecting user proportion information, content popularity distribution information, cache capacity information and link delay information in a satellite-ground cooperative network; on the basis of a differentiated user interest and differentiated access mode, determining a delay optimization-oriented base station and satellite cooperative content caching problem; solving the content caching problem to obtain a base station and satellite cooperative content caching strategy; and caching content in a base station and a satellite on the basis of the obtained content caching strategy, and providing a content service for base station access users and satellite access users in the network. According to the method, a content service delay for base station access users and satellite access users can be reduced at the same time under differentiated user interest distribution, the problem of a long communication delay caused by a long transmission distance of a satellite-ground link is relieved, and the user experience is improved, so that various delay sensitive services are supported.
A dynamic face specular material extraction algorithm based on a time-varying light field of a lightweight device, comprising the following steps: (1) according to an open-source face skin reflection parameter statistical database Merl/ETH Skin, expressing face specularity by a Torrance-Sparrow model; (2) converting the Torrance-Sparrow model of the skin into a Rusinkiewicz Half-vector parameterization expression mode; (3) solving for a normal by means of at least three included angles between the normal and different Half-vectors; and (4) after obtaining the normal, substituting the normal into the Torrance-Sparrow model to calculate the reflection intensity of the specularity. The present invention further provides a device for a dynamic face specular material extraction algorithm based on a time-varying light field of a lightweight device. According to the present invention, a face skin material reflection model is fully utilized, and an extracted normal has a physical basis; the present invention can achieve extraction of specularity parameters by means of not more than three illumination modes, achieves less time interval, and is suitable for a face dynamic scenario.
A lightweight dynamic human face acquisition apparatus based on a time sequence varying light field, comprising a camera array and a time sequence controllable light field array. The light field array provides illumination for the camera array. The camera array and light fields are distributed around the front human face, are presented in an approximately semicircular shape in a top view, and are distributed in multiple layers in a vertical direction. The camera array and LED light sources are connected to a control module, so that the photographing of cameras and the on/off of the light sources can be synchronously controlled, and a time sequence varying light field is formed in a dynamic human face acquisition process. The present invention further provides an acquisition method of the lightweight dynamic human face acquisition apparatus based on a time sequence varying light field. According to the present invention, by means of a multi-layer LED light field array, a set of illumination modes with uniform skin diffuse reflection components and different high-light components are designed, so that a dynamic scene-friendly time sequence varying light field is realized, and the effect of dynamic extraction of geometry and human face physical materials is ensured. The present apparatus achieves high-quality reconstruction of human face geometry and materials at relatively low hardware cost.
The present disclosure discloses a method and apparatus for monitoring a link state between nodes. The method includes: determining, by a first node, whether to transmit a first diagnosis message to a second node based on a monitoring result of whether to receive message data transmitted by the second node for a first time duration; processing, by the second node, the received first diagnosis message and transmitting the processed second diagnosis message to a third node; processing, by the third node, the received second diagnosis message and transmitting the processed third diagnosis message to the second node; determining, by the second node, whether to transmit the third diagnosis message to the first node; and determining, by the first node, a link state between the first node and the second node based on a monitoring result of whether to receive the third diagnosis message for a second time duration.
H04L 43/0811 - Surveillance ou test en fonction de métriques spécifiques, p. ex. la qualité du service [QoS], la consommation d’énergie ou les paramètres environnementaux en vérifiant la disponibilité en vérifiant la connectivité
H04L 12/437 - Isolement de la défaillance de l'anneau ou reconfiguration
A surgical robot and a surgical system. The surgical robot is used for manipulating a flexible ureteroscope (90) and a holmium laser fiber (92). The flexible ureteroscope (90) comprises a working channel used for accommodating the holmium laser fiber (92). The surgical robot comprises a mounting base (70), and a fixing assembly (71), a guiding assembly (81) and a twisting and conveying assembly (80) which are arranged on the mounting base. The fixing assembly (71) is configured to fixedly mount the flexible ureteroscope (90) on the mounting base (70). The guiding assembly (81) is located between the twisting and conveying assembly (80) and the fixing assembly (71) and is communicated with the working channel of the flexible ureteroscope (90). The twisting and conveying assembly (80) is configured to clamp and drive the holmium laser fiber (92), so that the holmium laser fiber (92) passes through the guiding assembly (81) and then reciprocates in the working channel of the flexible ureteroscope (90).
A61B 18/26 - Instruments, dispositifs ou procédés chirurgicaux pour transférer des formes non mécaniques d'énergie vers le corps ou à partir de celui-ci par application de radiations électromagnétiques, p. ex. de micro-ondes en utilisant des lasers le faisceau étant dirigé le long, ou à l'intérieur d'un conduit flexible, p. ex. d'une fibre optiquePièces à main à cet effet pour produire une onde de choc, p. ex. lithotritie par laser
The present disclosure relates to a bending mechanism and a surgical robot. The bending mechanism comprises a mounting base for fixing a flexible ureteroscope, a rotating disc mounted on the mounting base, a bending driving member mounted on the mounting base, and a connection member arranged on the rotating disc. The connection member is used for being connected to a handle. The bending driving member is used for driving the rotating disc to rotate, such that the connection member rotates the handle. In the present disclosure, the bending driving member drives the rotating disc to rotate, so as to drive the handle of the flexible ureteroscope to rotate, such that the flexible ureteroscope can freely bend within the travelling range thereof and accurately stop at any certain angle within the travelling range. The bending mechanism and the surgical robot have simple structures and low costs.
The present disclosure provides a data processing method and apparatus, a storage medium and an electronic device. In the method, after receiving data frames, a switching chip using a data analysis model deployed in a data processing unit to analyze the data frames; according to an analysis result, determining a processing policy for the data frames; and, according to the processing policy, the switching chip processing the data frames.
The present application relates to a knowledge graph entity linking method and apparatus, and a computer device and a storage medium. The method comprises: on the basis of a question sample, an entity mention sample, a knowledge graph entity positive sample and a knowledge graph entity adjacency sub-graph sample, acquiring a training data positive sample; on the basis of the question sample, the entity mention sample, a knowledge graph entity negative sample and a corresponding knowledge graph entity adjacency sub-graph sample, acquiring a training data negative sample; training an entity linking initial model on the basis of the training data positive sample and the training data negative sample, so as to obtain an entity linking model; and inputting a user question, an entity mention, a candidate knowledge graph entity and a corresponding knowledge graph entity adjacency sub-graph into the trained entity linking model, so as to determine a target knowledge graph entity linked with the entity mention. The problems existing in the prior art of the effect of an entity consistency model being poor and the entity linking accuracy being relatively low in a question-answering scenario are solved.
G06F 16/955 - Recherche dans le Web utilisant des identifiants d’information, p. ex. des localisateurs uniformisés de ressources [uniform resource locators - URL]
98.
DETERMINISTIC DATA TRANSMISSION DEVICE AND METHOD FOR COMPATIBILITY NETWORK
The present disclosure relates to a deterministic data transmission device and method for a compatibility network. The device includes a data link layer and a physical layer of an open system interconnection. The data link layer includes a support module of preemption function of a MAC client, a MAC control module, and a MAC merge module, and the physical layer includes a coordination sublayer and a port physical layer. And the support module of preemption function of the MAC client is connected to the MAC control module, the MAC merge module is connected to the coordination sublayer, and the port physical layer includes a physical transport medium connected to the coordination sublayer. The deterministic data transmission method for a compatibility network includes slicing of low priority data, enveloping low priority frames, combining sliced frames, and verifying a combination of sliced frames.
The present disclosure discloses a time-sensitive network switch including a plurality of multi-core CPUs. In the time-sensitive network switch, by the parallel data processing method with multi-core system-on-chips (SoC), industrial real-time data are distributed to multi-core SoCs for processing in parallel. The processed data are scheduled according to the identified priority, and are arranged to the different priority queues of the port in order to process the high-priority data first, which reduces the processing time for the data in the device. The data passes through the security encryption engine in the time-sensitive network switch, which ensures the security for processing data in the device. The reliability, real-time and stability of the data transmission in the time-sensitive network are improved.
G06F 13/28 - Gestion de demandes d'interconnexion ou de transfert pour l'accès au bus d'entrée/sortie utilisant le transfert par rafale, p. ex. acces direct à la mémoire, vol de cycle
Disclosed in the present disclosure are an integrated fast data exchange method and a time-sensitive network switch. A network switching device performs sensitivity recognition on data to be forwarded; for exchange and transmission of non-time-sensitive data, a push port queue and a pop port queue perform cut-through forwarding, reducing the retention time of a data packet in the switching device; and for time-sensitive data, the push port queue and the pop port queue perform data forwarding according to a corresponding priority scheduling strategy. According to the present disclosure, fast forwarding processing of different types of data frames is realized by means of different types of data frame exchange strategies in a same apparatus, so that the problem of the industrial production efficiency being affected by long processing time for time-sensitive data in a switching apparatus caused by conventional store-and-forward modes having a large delay in data processing and performing error detection on a data packet entering a switch can be solved.
