Product
Patsnap Eureka
For R&D, Patsnap Eureka makes reading and utilizing patents & technical documents easy.
Patsnap Eureka AIR
Designed for self-driven R&D workflows. Generate viable solutions, solve complex R&D challenges, empower your innovation with AI.
Patsnap Eureka Materials
Designed for material experts only. Revolutionize your material R&D, from search, analyze, to developing new materials.
Feature
TechResearch
Generate reliable direction feasibility study reports for your R&D in just a few steps.
TechSeek
Discover and master advanced knowledge NOW. Basics, ideas, possibilities, all at once.
TechMind
As an expert in R&D Theories, TechMind can generates customized viable solutions instantly.
TechRisk
Analyze your overall solution with one click, know your potential R&D risks in advance.
TechMonitor
Get weekly tech updates, stay abreast of the latest tech innovations and key insights.
Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
32results about How to "Ensure training efficiency" patented technology
Efficacy Topic
Property
Owner
Technical Advancement
Application Domain
Technology Topic
Technology Field Word
Patent Country/Region
Patent Type
Patent Status
Application Year
Inventor
Mobile ad hoc network intrusion detection method and device based on deep learning
ActiveCN104935600AEnsure training efficiencyGuaranteed detection efficiencyTransmissionSecurity arrangementData acquisitionWireless data
The invention discloses a mobile ad hoc network intrusion detection method and device based on deep learning, relating to the field of wireless network safety. The device comprises a data acquisition module, a data fusion module, a preprocessing module, a storage module, an intrusion detection module and a response warning module. After fusion and redundancy elimination of captured wireless data packages, network behavior features are extracted and stored; after deep learning of the network behavior features, a deep neural network model expressing network behaviors is established; and to-be-detected network data is input into the deep neural network model, after intrusion is judged and recognized, response and warning are performed. According to the method, network behavior feature vectors which are detected and are considered to be abnormal are stored and are used for training the deep neutral network. When occurring again, the intrusion types can be detected and recognized. While the model training and detection efficiency are guaranteed, the detection accuracy is improved, and the safety of the mobile ad hoc network is further improved.
Owner:NO 54 INST OF CHINA ELECTRONICS SCI & TECH GRP
Attention mechanism and hash-based image retrieval method and device and storage medium
ActiveCN111723220AHigh expressionEnhance expressive abilityStill image data indexingStill image data queryingFeature extractionOriginal data
The invention discloses an attention mechanism and hash-based image retrieval method and device and a storage medium, and the method comprises the steps: obtaining the discrete hash code of each imagebased on a pre-trained attention mechanism-based hash feature extraction model for each image in an original data set, and building an image feature library; obtaining a discrete hash code of a to-be-detected image based on a pre-trained hash feature extraction model based on the attention mechanism; and querying data closest to the discrete hash code Hamming distance of the to-be-detected imagein an image feature library, wherein the image corresponding to the data is a retrieval result. The Hash feature extraction model based on the attention mechanism gives full play to the ability of thedeep convolutional neural network to extract the abstract semantic features of the image; and by embedding the attention module, the network can focus on the visual content capable of expressing semantic information of the whole image as much as possible in the image, the expression effect of hash coding is improved, and image retrieval is more accurate and faster.
Owner:CENT SOUTH UNIV
Image classification method of spiking learning model based on dynamic threshold
InactiveCN111046954AEnsure training efficiencyGuaranteed accuracyCharacter and pattern recognitionData setClassification methods
The invention discloses an image classification method of a spiking learning model based on a dynamic threshold. The method comprises the following steps: obtaining an image data set, converting imageinformation into a pulse excitation sequence by adopting a phase delay coding method, establishing and training a spiking learning model based on a dynamic threshold, and performing classification processing on images to be classified by utilizing the trained spiking learning model. According to the invention, image information is converted into a pulse excitation sequence by using a phase delaycoding method; according to the image classification method and device, the dynamic threshold is set up, the spiking learning model based on the dynamic threshold is set up for training, the trained spiking learning model is used for image classification, the robustness of the learning model can be remarkably improved while the training efficiency and accuracy of the learning model are guaranteed,and the image classification efficiency is further improved.
Owner:UNIV OF ELECTRONICS SCI & TECH OF CHINA
Hand fine action training guidance system and method based on vision
InactiveCN111145865AReal-time detection of fine movementsImprove rehabilitation effectPhysical therapies and activitiesNeural architecturesPhysical medicine and rehabilitationData profiling
The invention relates to a hand fine motion training guidance system and method based on vision. The system comprises a data processing module, a 3D hand posture recognition module, a data analysis module and a training guidance module. The system detects the hand fine movement ability of a user in real time through a computer vision method, guides the user to perform hand fine movement rehabilitation training in real time, guarantees the training efficiency and intensity, accurately controls training parameters, and objectively evaluates the rehabilitation training effect through quantitativedata.
Owner:HEFEI INSTITUTES OF PHYSICAL SCIENCE - CHINESE ACAD OF SCI
Offshore wind power ultra-short-term prediction method based on LSTM deep learning network
InactiveCN110276472APrevent overfittingEnsure training efficiencyForecastingNeural architecturesOverfittingLearning network
The invention discloses an offshore wind power ultra-short-term prediction method based on an LSTM deep learning network. The offshore wind power ultra-short-term prediction method comprises the following steps: firstly, carrying out various analysis on a wind speed time sequence, and combining various analysis results to extract the characteristic quantity of offshore wind power ultra-short-term prediction, thereby avoiding overfitting of a model caused by excessive input quantity; carrying out modeling prediction on the offshore wind speed by utilizing a deep learning method; and combining the actual wind speed-power curve of the offshore wind generating set to obtain an offshore wind power ultra-short-term predicted value. According to the offshore wind power ultra-short-term prediction method, the characteristic quantity of the ultra-short-term prediction of the offshore wind power is extracted according to the characteristics of the offshore wind speed, so that the input of the characteristic quantity with low correlation is avoided, and the prediction efficiency is improved while the offshore wind power prediction precision is improved.
Owner:SOUTHERN POWER GRID PEAK LOAD & FREQUENCY REGULATION GENERATING CO LTD +1
Text generation method and device
ActiveCN111563375AEnsure training efficiencyGuaranteed recognition accuracySemantic analysisFeature vectorTarget text
The invention provides a text generation method and device, and relates to the technical field of natural languages. The invention provides a text generation method and device. The method comprises the steps of determining a first segmented word in a first text corpus; under the condition that the first segmented word is matched with the preset feature, replacing the first segmented word with a feature mark corresponding to a preset feature to obtain a second text corpus, obtaining a first word vector corresponding to a first segmented word in the first text corpus, the first segmented word and the first feature vector corresponding to the feature mark in the second text corpus; inputting the first word vector and the first feature vector into the text generation model at the moment, obtaining the output target word vectors of the target semantic structure, and then obtaining the target text corpus according to the target word vectors. The target text corpus obtained by the embodimentof the invention comprises the required and complete target semantic structure, and the first text corpus does not limit the obtaining mode, so that the problems of high similarity and poor extensioneffect between the extended target text corpuses are avoided.
Owner:BEIJING SINOVOICE TECH CO LTD
Image defogging method and system based on cyclic generative adversarial network
ActiveCN113658051AEliminate requirementsSolve the problem that cannot be applied to real dehazing scenesImage enhancementImage analysisData setAlgorithm
The invention discloses an image defogging method and system based on a cyclic generative adversarial network. The method comprises the steps of obtaining a to-be-processed foggy image; inputting the image into a pre-trained dense connection cyclic generative adversarial network, and outputting a fog-free image, wherein the dense connection cyclic generative adversarial network comprises a generator comprising an encoder, a converter and a decoder, the encoder comprises a dense connection layer for extracting features of an input image, the converter comprises an over-conversion layer for combining the features extracted by the encoder stages, the decoder comprises a dense connection layer and a scaled convolutional neural network layer, the dense connection layer is used for restoring the original features of the image, the scaled convolutional neural network layer is used for removing the checkerboard effect of the restored original features, and a finally output fogless image is obtained. The method and system have the advantages that image defogging is carried out based on the cyclic generative adversarial network, the requirement for a pairwise data set is eliminated, the utilization rate of the feature map is improved, the network training efficiency is kept, and the quality of the generated image is improved.
Owner:NANJING UNIV OF POSTS & TELECOMM
Network flow time sequence prediction method based on distributed clustering
ActiveCN107067028APrediction is accurateOvercome deficienciesCharacter and pattern recognitionData switching networksTraffic capacityCluster algorithm
The invention discloses a network flow time sequence prediction method based on distributed clustering. The method comprises the steps: carrying out the fusion of a distributed clustering algorithm and an auto-regression model; obtaining time segment groups through the segmenting of time sequence data, and carrying out the distributed clustering of the time segment groups through employing a distributed K-means clustering algorithm; carrying out the normal distribution fitting of each cluster in a clustering result, and obtaining a normal distribution N(mu, sigma2), wherein mu is a preliminary prediction value. At a prediction stage, the to-be-predicted time sequence data is preprocessed, and the to-be-predicted time segment prefix groups. The method achieves the distributed calculation of the preliminary prediction value corresponding to the nearest cluster, and achieves the correction of the preliminary prediction value through combining with an autoregressive model, thereby obtaining a more precise final prediction value.
Owner:SOUTH CHINA UNIV OF TECH
Training corpus generation method of intention recognition model and related equipment thereof
ActiveCN112395390AImprove accuracySolve the problem of not filling the query categoryCharacter and pattern recognitionEnergy efficient computingNatural language processingMedicine
The embodiment of the invention belongs to the field of big data, is applied to the field of smart medical treatment, and relates to a training corpus generation method of an intention recognition model and related equipment of the training corpus generation method. Wherein the customer answer corpus comprises an inquiry related corpus and a non-inquiry related corpus; establishing an inquiry related corpus and a non-inquiry related corpus; adjusting the inquiry-related corpus and the non-inquiry-related corpus based on the similarity between the non-inquiry-related corpus and the inquiry-related corpus to obtain a target inquiry-related corpus and a target non-inquiry-related corpus; establishing a first training sample based on the target inquiry related corpus; establishing a second training sample based on the intention label and a non-inquiry related corpus; and taking the first training sample and the second training sample as training corpora and outputting the training corpora.Wherein the training corpus can be stored in a block chain. The quality of the training corpus is improved.
Owner:PING AN TECH (SHENZHEN) CO LTD
Federated learning training method and device for high-delay network environment robustness
ActiveCN112001502AEnsure training efficiencyEnsemble learningData switching networksSimulationIndustrial engineering
The invention discloses a federated learning training method and device for high-delay network environment robustness, computer equipment and a storage medium, and relates to the artificial intelligence technology. The method comprises: obtaining the current system time, and obtaining a corresponding target data uploading terminal if the encrypted data uploaded by a plurality of data uploading terminals is not received; obtaining a current network time delay value of each target data uploading terminal to obtain a maximum network time delay value; calculating to obtain a delay step length according to the maximum network delay value and the unit time sequence interval step length; summing the current system time and the delay step length to obtain target system time; and if the current time is the target system time and the target encrypted data uploaded by the target data uploading terminal is not received, stopping the local federation learning training until the target encrypted data uploaded by all the target data uploading terminals is received, and recovering the local federation learning training. Under the condition of network delay, the training efficiency of federated learning is kept in a time delay sparse updating mode.
Owner:PING AN TECH (SHENZHEN) CO LTD
VR-based teaching and training system
PendingCN108735031AImprove answering efficiencyImprove accuracyCosmonautic condition simulationsElectrical appliancesVirtual teachingDependability
The invention relates to the technical field of virtual teaching and in particular to a VR-based teaching and training system comprising at least one group of teaching units for storing teaching documents, a classification unit for classifying data in the teaching units, and a plurality of learning units connected with the classification unit to identify the teaching data classified by the classification unit, wherein each learning unit is provided therein with a question feedback unit, the classification unit is provided therein with a feedback receiving unit, the feedback receiving unit includes an automatic feedback question module and a manual solution module which transmits the feedback questions to the teaching units. The VR-based teaching and training system automatically records and answers the training questions, can record the questions to prevent the same questions from being required to be manually answered, can automatically answer the feedback questions, and has high solution efficiency, high accuracy and high reliability.
Owner:深圳市辰展教育科技有限公司
Pelvis correction rehabilitation training robot rehabilitation training control method
ActiveCN111449899AEnsure training efficiencySimple structureChiropractic devicesMuscle exercising devicesMuscle strengthElectric machinery
The invention discloses a pelvis correction rehabilitation training robot rehabilitation training control method comprising the following steps: step 1, before rehabilitation training, using a pressure sensor to collect the maximum muscle force value of a patient, and using a photoelectric encoder to collect the maximum motion angle value of the hip joint of the patient; step 2, setting a maximummuscle strength value and a maximum movement angle value of a hip joint, repeatedly outputting motor torque and rotating speed in a range smaller than the maximum muscle strength value and the maximummovement angle value of the hip joint, measuring the muscle strength of the patient and the movement angle of the hip joint, giving a pelvic deformation condition evaluation result and rehabilitationtraining data of the patient, and generating an evaluation report; and step 3, setting rehabilitation training modes and parameters according to the evaluation report, generating a control instruction, completing setting of motor torque and rotating speed, and enabling the motor to drive the patient to complete corresponding actions. A servo motor is adopted to drive flexion and extension movement of hip joints of a robot, and the robot is used for replacing a treatment practitioner to assist a patient in rehabilitation training of pelvis correction.
Owner:ZHEJIANG UNIV
A mobile ad hoc network intrusion detection method and device based on deep learning
ActiveCN104935600BEnsure training efficiencyGuaranteed detection efficiencyTransmissionSecurity arrangementPattern recognitionData pack
The invention discloses a mobile ad hoc network intrusion detection method and equipment based on deep learning, and relates to the field of wireless network security. The device of the present invention includes a data acquisition module, a data fusion module, a preprocessing module, a storage module, an intrusion detection module, and a response alarm module, and after the captured wireless data packets are fused and de-redundant, network behavior characteristics are extracted and stored; After in-depth learning of network behavior characteristics, a deep neural network model is established to express network behavior; the network data to be detected is input into the deep neural network model, and the intrusion judgment and identification are completed to respond to the alarm. The method of the invention stores the detected abnormal network behavior feature vectors and uses them to train the deep neural network, and when these intrusion types occur again, they can be detected and identified. On the premise of ensuring model training and detection efficiency, the invention improves the detection accuracy and further improves the security of the mobile ad hoc network.
Owner:NO 54 INST OF CHINA ELECTRONICS SCI & TECH GRP
An English translation practice device
ActiveCN111583735BEfficient removalImprove purification effectCombination devicesElectrical appliancesActivated carbonDust control
The invention discloses an English translation practice device, which belongs to the field of English translation. An English translation practice device is provided with activated carbon adsorption materials inside the first adsorption panel and the second adsorption panel, which can effectively remove impurities in the air. Under the action of the purification panel, it is convenient to purify the air, and it is convenient and efficient to purify the air in the room, which greatly guarantees the user's living environment. Under the action of the blower, the purified air is discharged through the first outlet pipe respectively. To the room, it is convenient to play a good effect of dust removal. Through the second outlet pipe, it is convenient to play a refreshing role. By adjusting the degree of engagement between the panel monomers, and under the action of the curved connecting pipe, the The volatilized gas of the mint liquid is blown to the user's face, which can easily play a good refreshing effect and greatly ensure the training efficiency.
Owner:HUNAN UNIV OF ARTS & SCI
Neural network training method and device, storage medium and electronic device
ActiveCN109726808AEnsure training efficiencyEasy to handleNeural architecturesNeural learning methodsHuman–robot interactionData mining
The invention discloses a neural network training method and device, a storage medium and an electronic device. The method comprises the steps that an offline sample set used for training a neural network in a human-computer interaction application is acquired, and the offline sample set comprises offline samples meeting preset configuration conditions; The initial neural network is trained offline by using the offline sample set to obtain an object neural network, and in the human-computer interaction application, the processing capability of the object neural network is higher than that of the initial neural network; And the object neural network is accessed to an online operation environment of the human-computer interaction application for online training to obtain a target neural network. According to the invention, the technical problem of low training efficiency in the neural network training method provided by the related technology is solved.
Owner:TENCENT TECH (SHENZHEN) CO LTD
Pelvis correction rehabilitation training robot
ActiveCN111449900AEnsure training efficiencySimple structureChiropractic devicesNursing bedsElectric machineryEngineering
The invention discloses a pelvis correction rehabilitation training robot. The robot comprises a supporting body, a hip joint correction module and a leg supporting module, wherein the hip joint correction module comprises a motor, a speed reducer and a hip joint correction part; the motor is installed on the supporting body; an output shaft of the motor is connected with the input end of the speed reducer; the output end of the speed reducer is connected with the hip joint correction part; and the hip joint correction part is connected with the leg supporting module. A servo motor is adoptedto drive flexion and extension movement of hip joints of the robot, and the robot is used for replacing a therapist to assist a patient in rehabilitation training of pelvis correction, so that the therapist can be liberated from a heavy training process, the training efficiency can be guaranteed, and the rehabilitation training robot has the great significance.
Owner:ZHEJIANG UNIV
Neural network model adjustment method, apparatus and device, and storage medium
PendingCN112434552AImprove generalization abilityAvoid forgettingCharacter and pattern recognitionNeural architecturesEngineeringNetwork model
The embodiment of the invention discloses a neural network model adjustment method and device, equipment and a storage medium, and relates to the technical field of neural networks, and the method comprises the steps: inputting a target number of sample data into a neural network model, and updating the model parameters of the neural network model according to a first output result of the neural network model, wherein the model parameter before updating is a first parameter, and the model parameter after updating is a second parameter; and inputting the target number of application data into the updated neural network model, and correcting model parameters of the neural network model according to a second output result and a second parameter of the updated neural network model, the model parameters before correction being the first parameter, and the model parameters after correction being the third parameter. By adopting the method, the technical problem that the performance of the deep learning model cannot be ensured when the deep learning model is applied to different scenes in the prior art can be solved.
Owner:GUANGZHOU SHIYUAN ELECTRONICS CO LTD
Mammary gland X-ray image antisymmetric generation analysis model training method and device
PendingCN111415333AImprove analysis accuracyEnsure training efficiencyImage enhancementImage analysisLeft breastFeature extraction
The embodiment of the invention provides a mammary gland X-ray image antisymmetric generation analysis model training method and device. The problems that an existing mammary gland X-ray image analysis mode is low in efficiency and accuracy are solved. The mammary gland X-ray image antisymmetric generation analysis model training method comprises: respectively extracting original bilateral mammarygland image features of bilateral mammary gland X-ray images based on a pre-trained feature extraction model, wherein the bilateral mammary gland X-ray images correspond to a left breast and a rightbreast respectively; inputting the original bilateral mammary gland image features into a neural network model to obtain an antisymmetric analysis result opposite to the feature information of the original bilateral mammary gland image features; calculating a loss function value between the generated antisymmetric analysis result and original bilateral mammary gland image features; and adjusting parameters of the neural network model based on the loss function value.
Owner:BEIJING SHENRUI BOLIAN TECH CO LTD +1
Training progress compensation method and device based on middle platform and middle platform
ActiveCN113409033AEasy to learnIn line with learning rulesOffice automationSimulationIndustrial engineering
The invention provides a middle desk-based training progress compensation method and device and a middle desk, and the middle desk obtains training video data of a class recording end in real time and executes the following operations: acquiring current machine information of an in-service worker; when the current machine information of any on-duty worker at the first moment meets a first preset state, performing first intervention marking at the first moment of the training video data; when the current machine information of the in-service staff at the second moment is changed into a second preset state from the first preset state, performing second intervention marking at the second moment of the training video data; based on an offset interception request, performing time offset processing on the first intervention mark and the second intervention mark to obtain a first new compensation mark and a second new compensation mark; and intercepting the training video data based on the first new compensation mark and the second new compensation mark to obtain a training progress compensation video, and sending the training progress compensation video to the corresponding on-duty worker.
Owner:STATE GRID ZHEJIANG ELECTRIC POWER +1
Exoskeleton-type upper limb rehabilitation robot
InactiveCN101829003BComfortable to useScientific upper limb rehabilitation trainingChiropractic devicesKinematicsShoulder Blades
The invention relates to a rehabilitation robot, in particular to an exoskeleton-type upper limb rehabilitation robot, which realizes the upper limb rehabilitation by assisting shoulder blades of the body in rotation. The exoskeleton-type upper limb rehabilitation robot comprises an upper arm part, an upper arm forward bending assisting part, an upper arm external expansion assisting part, a shoulder blade rotation assisting part, a transmission part and a bracket part and is characterized in that the diameter of a forward bending driving wheel of the transmission part is twice that of a driving wheel; the diameter of an external expansion driving wheel is twice that of the driving wheel; and a weight wheel, the forward bending driving wheel and the external expansion driving wheel are all provided with positioning devices. The exoskeleton-type upper limb rehabilitation robot follows a 2-1 principle of driving shoulder blades by the upper limb in human body articular kinesiology, makes the mechanical upper arms assist the shoulder blades in rotation according to the 2-1 principle when assisting the upper arms of the user in bending forward or extending by assisting the shoulder blades of the human body and can more scientifically carry out rehabilitation training of the upper limb of a patient.
Owner:青岛思威机器人科技有限公司
A table tennis ball collection device for table tennis training and its use method
InactiveCN111298394BAvoid being trampledAvoid the phenomenon of increased labor intensityBall sportsRacket sportsSimulationMechanical engineering
The invention relates to the technical field of table tennis training, and specially relates to the ping-pong ball collecting device for ping-pong ball training and the using method thereof. The device comprises a table, an adjusting device, a collecting device and a guiding device; an adjusting device is arranged at the bottom end of the table; the table is fixedly connected with the adjusting device; a collecting device is arranged at the top end of the adjusting device; the collecting device is fixedly connected with the adjusting device; a ball collecting hopper is arranged at the right end of the collecting device; in the present invention, the limiting cloth, the first fan and the second fan are arranged; the arrangement is matched with the fixed connection of the limiting cloth andthe support plate, the elastic force of the second spring to the extrusion plate, the fixed connection of the extrusion plate and the pressure sensor, the traction of the first fan to wind and the traction of the second fan to wind; when the device is used, the emitted table tennis balls can be automatically guided and collected, so that the phenomenon that the labor intensity of operators is increased due to manual collection is avoided, and the phenomenon that the table tennis balls are trampled during collection is also avoided.
Owner:郑州财税金融职业学院
Network Traffic Time Series Forecasting Method Based on Distributed Clustering
ActiveCN107067028BPrediction is accurateOvercome deficienciesCharacter and pattern recognitionData switching networksCluster algorithmAlgorithm
Owner:SOUTH CHINA UNIV OF TECH
A pelvic correction rehabilitation training robot for rehabilitation training
ActiveCN111449899BEnsure training efficiencySimple structureChiropractic devicesMuscle exercising devicesEvaluation resultMuscle strength
Owner:ZHEJIANG UNIV
A brain-computer interface decoding method based on spiking neural network
ActiveCN113298242BAchieving decoding accuracySmall amount of calculationNeural architecturesMachine learningLiquid stateSimulation
The invention discloses a brain-computer interface decoding method based on an impulse neural network, comprising: (1) constructing a liquid state machine model based on an impulse neural network, and the liquid state machine model is composed of an input layer, an intermediate layer and an output layer; wherein, The connection weight from the input layer to the intermediate layer is W hx , the weight of the recurrent connection inside the middle layer is W hh , the read weight from the middle layer to the output layer is W yh ; (2) Input the pulse EEG signal, and train each weight using the following strategies: (2-1) Adopt STDP unsupervised training to connect the weight W hx ; (2-2) Set the intermediate layer cyclic connection weight W by means of distance model and random connection hh ; (2‑3) Using Ridge Regression to train the connection weights W with supervised supervision yh , and establish the mapping between the intermediate layer liquid information R(t) and the output motion information Y(t), and finally output the predicted motion trajectory. By using the present invention, the model can be quickly trained in a relatively short time, the movement trajectory of the arm can be predicted in real time, and the efficiency and accuracy can be improved.
Owner:ZHEJIANG UNIV
Robust federated learning training method and device in high-latency network environment
ActiveCN112001502BEnsure training efficiencyEnsemble learningSecuring communicationDistributed computingComputer engineering
The invention discloses a robust federated learning training method, device, computer equipment and storage medium in a high-delay network environment, and relates to artificial intelligence technology, including obtaining the current system time, and obtaining the corresponding The target data upload terminal; obtain the current network delay value of each target data upload terminal to obtain the maximum network delay value; calculate the delay step according to the maximum network delay value and the unit timing interval step; the current system time The target system time is obtained by summing the delay step length. If the current time is the target system time and the target encrypted data uploaded by the target data upload terminal has not been received, the local federated learning training will be stopped until all target data upload terminals are received. After the target encrypts the data, resume the local federated learning training. In the case of network delay, the training efficiency of federated learning is maintained through delay sparse update.
Owner:PING AN TECH (SHENZHEN) CO LTD
A pelvic correction rehabilitation training robot
ActiveCN111449900BEnsure training efficiencySimple structureChiropractic devicesNursing bedsPelvic regionPhysical medicine and rehabilitation
The invention discloses a pelvic correction rehabilitation training robot, comprising a support body, a hip joint correction module and a leg support module, the hip joint correction module includes a motor, a reducer and a hip joint correction piece, the motor is installed on the support body, and the motor output shaft It is connected with the input end of the reducer, the output end of the reducer is connected with the hip joint correction part, and the hip joint correction part is connected with the leg support module. Using servo motors to drive the flexion and extension of the robot's hip joints, and using robots to replace therapists to assist patients in pelvic correction rehabilitation training can not only liberate therapists from the heavy training process, but also ensure training efficiency, which is of great significance.
Owner:ZHEJIANG UNIV
A Knowledge Graph Fusion Method Based on Entity Alignment
ActiveCN110941722BImprove alignment resultsEasy alignmentCharacter and pattern recognitionNeural architecturesKnowledge graphData mining
The invention discloses a knowledge map fusion method based on entity alignment, comprising the following steps: acquiring data of two knowledge maps; learning the structure vector of the entity by using a graph convolution network, and expressing the name of the entity as a word vector; calculating the data between the entities The comprehensive distance is used to represent the similarity between entities; an iterative training framework based on curriculum learning is used for entity recognition and alignment; according to the results of entity alignment, two knowledge maps are merged into one knowledge map. The method of the invention designs a basic framework of entity alignment that integrates structural features and entity name features; designs an iterative training method based on curriculum learning, expands training data from easy to difficult, and uses a word-shift distance model to repeat the pre-order alignment results. Sorting to fully mine entity name information, making the integration of knowledge graphs more accurate and comprehensive.
Owner:NAT UNIV OF DEFENSE TECH
Marine rescue device and intelligent rescue system
PendingCN114771774AImprove accuracyIncrease rescue rateCast linesLocation information based serviceMarine engineeringLaunch craft
The invention discloses a marine rescue device and an intelligent rescue system. The invention relates to a maritime rescue device, which belongs to the field of maritime rescue, and comprises more than one intelligent bracelet with a waterproof function, a positioning module and a vital sign monitoring module, the lifesaving equipment can be automatically inflated after being ejected; the launcher comprises a rotary table, a vertical rod, a shooting barrel and a driving mechanism, the vertical rod is rotationally connected with the shooting barrel, a rotating wheel and a thrust structure are arranged in the shooting barrel, a rope is wound on the rotating wheel, one end of the rope is connected with the lifesaving equipment, and the thrust structure is used for launching the lifesaving equipment; the console comprises a microprocessor, and a storage module and a communication module which are electrically connected with the microprocessor, and the console is in wireless communication connection with the intelligent bracelet and is electrically connected with the driving mechanism.
Owner:湖南数军物联网科技有限公司
A middle platform-based training progress compensation method, device and middle platform
The present invention provides a training progress compensation method, device and middle platform based on the middle platform. The middle platform obtains the training video data of the recording terminal in real time and performs the following operations, including: obtaining the current machine information of the on-the-job staff; When the current machine information of the personnel at the first moment meets the first preset state, the first intervention mark is performed at the first moment of the training video data; when the current machine information of the in-service staff at the second moment is changed from the first preset state When it becomes the second preset state, the second intervention mark is performed at the second moment of the training video data; based on the offset interception request, the time offset processing corresponding to the first intervention mark and the second intervention mark is obtained to obtain the first new A compensation mark and a second new compensation mark; intercepting the training video data based on the first new compensation mark and the second new compensation mark to obtain a training progress compensation video, and sending the training progress compensation video to the corresponding in-service staff .
Owner:STATE GRID ZHEJIANG ELECTRIC POWER +1
Coarse-grained soil filler gradation recognition method and application system based on convolutional neural network
ActiveCN111523616BEnsure training efficiencyImprove accuracyCharacter and pattern recognitionNeural architecturesSoil scienceImaging processing
The invention relates to a coarse-grained soil filler gradation recognition method and application system based on a convolutional neural network, constructing a first neural network and several second neural networks; collecting images of fillers on site, and using the first neural network to output the classification of particle size ranges Results; input the filler image and classification results into the second neural network corresponding to the particle size range to obtain the quality of coarse-grained soil in each single particle size range; count the mass distribution of coarse-grained soil in various particle size ranges to obtain gradation and grade matching curve. The present invention obtains the particle size range through the first neural network, and obtains the total mass of a single particle size group through the corresponding second neural network, thereby improving the accuracy of mass calculation, thereby ensuring the accuracy of gradation calculation. Multiple second neural networks are processed in parallel, which ensures the training efficiency of the second neural network and the on-site processing efficiency. High degree of automation, no need for complex image processing algorithms, no manual intervention, no dependence on operator experience, strong environmental adaptability, and high precision.
Owner:RAILWAY ENG RES INST CHINA ACADEMY OF RAILWAY SCI +2
Who we serve
- R&D Engineer
- R&D Manager
- IP Professional
Why Patsnap Eureka
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com