39results about How to "Accelerated training" patented technology

The invention relates to an underwater acoustic target recognition system and method based on multi-modal depth feature fusion, and belongs to the field of underwater acoustic target recognition. The method comprises the following steps: firstly, performing preprocessing on an obtained underwater acoustic target original audio and associated metadata, and constructing a multi-modal data set; and then the constructed multi-modal sample is input into an identification model, the model extracts deep representation of each modal through a multi-branch feature coding network, depth alignment and complementary aggregation of different modal features are realized by using a cross-modal cross attention mechanism guided by potential query, and a target identification result is output based on a decision network of mixed experts. According to the method, experimental verification is carried out on two disclosed underwater acoustic data sets, the experimental result verifies the effectiveness of the multi-modal deep fusion and hybrid expert adaptive decision strategy adopted by the method, and through mining the complementary advantages of acoustic features and semantic priori, the multi-modal deep fusion and hybrid expert adaptive decision strategy is obtained. And the robustness and generalization ability of the underwater acoustic target recognition system in the strong-noise and multi-working-condition environment are remarkably improved.

The invention relates to a credit distribution method in multi-agent cooperation, and the method comprises the following steps: obtaining joint data of a multi-agent system, the joint data comprising observation information and execution actions of each agent; the joint data are input into a discriminator model, and the discriminator model carries out modeling on an interaction dependency relationship among multiple agents based on an interaction dependency relationship modeling module of an attention mechanism; generating an auxiliary reward signal corresponding to each agent through a discriminator model; generating a fused reward signal based on the auxiliary reward signal and a global reward signal from the environment; and training and updating the multi-agent strategy model by using the fused reward signal so as to optimize the cooperation strategy of the multi-agent system. By introducing a credit distribution mechanism based on a discriminator model, automatic decomposition and optimization generation of individual reward signals are realized, so that the learning efficiency and stability of a multi-agent system in a complex cooperative task are remarkably improved.

The application discloses a low-latency computing task unloading method and system based on an artificial fish school algorithm, and comprises the following steps: establishing a MEC system for unloading computing tasks of multiple user devices to MEC servers; the MEC system comprises K MEC servers and N user devices; the user device computing task is completed through the pre-optimized MEC system, and the process comprises the following steps: dividing the computing task of each user device into K+1 subtasks, compressing part of the data of the subtasks by using a data compression method to obtain compressed data, dividing the subtasks into compressed data and uncompressed data, transmitting the K subtasks of the user device to each MEC server in a non-orthogonal multiple access mode, and respectively calculating the K+1 subtasks of the user device and each MEC server, and returning the calculation results of the MEC servers to the user device; the maximum user computing task completion latency is significantly reduced, and the user experience is improved.

This invention relates to an intelligent decision support system for rectal cancer prognosis integrating digital pathological images, belonging to the field of medical image processing technology. The system includes a data acquisition module for collecting multiple sets of sample data; a feature selection module for identifying multiple key medical imaging features of rectal cancer based on the multiple sets of sample data; a model building module for acquiring a pre-trained deep learning model and constructing a rectal cancer prognosis prediction model based on transfer learning, the pre-trained deep learning model, the multiple sets of sample data, and the multiple key medical imaging features of rectal cancer; an image acquisition module for acquiring pre- and post-operative medical images of the rectal cancer patient to be evaluated; and a recurrence prediction module for predicting the recurrence probability of the rectal cancer patient to be evaluated based on the pre- and post-operative medical images of the patient using the rectal cancer prognosis prediction model. This system has the advantage of improving the accuracy of non-invasive prediction of postoperative recurrence of rectal cancer.

The invention discloses a high-dimensional sparse text-oriented classification method, system and device and a storage medium, and belongs to the technical field of natural language processing. The method comprises the steps of obtaining to-be-classified text data; based on the to-be-classified text data, constructing a candidate redundant feature set by calculating semantic similarity; joint mutual information of candidate redundant features in the candidate redundant feature set is calculated, and key features are obtained through screening according to the joint mutual information; inputting the key features into a pre-constructed text classification model to obtain a classification prediction result; wherein the text classification model comprises a dual-channel gating module, an attention module and a prediction module; the two-channel gating module extracts context features in the key features; the attention module weights the context features to obtain weighted features; and the prediction module obtains a classification prediction result according to the weighted features. According to the method, effective compression of feature dimensions and precise modeling of context semantics are realized.

This invention discloses a method and system for improving the accuracy of target detection and classification in traffic scenes, comprising the following steps: S1: acquiring traffic video; S2: converting the traffic video into traffic images; S3: recognizing the traffic images based on a convolutional neural network to obtain the classification results of the traffic images; S4: issuing warnings based on the classification results of the traffic images. The beneficial effects of this invention are: it can improve the convergence speed of the model, accelerate the training of the model, and improve the recognition accuracy of the model.

The application provides a kind of double-link point cloud completion method, system, equipment and storage medium, method includes: the original point cloud data is respectively input into first feature extraction branch and second feature extraction branch, and first feature vector and second feature vector are respectively obtained by different network architecture;The first feature vector and the second feature vector are fused to obtain global high-dimensional semantic feature;And the global high-dimensional semantic feature is decoded to generate the target point cloud set of completion.This application uses two branches to simultaneously extract different dimensional features of point cloud, obtain more, richer high-dimensional semantic features and local features containing global point cloud information, which can improve point cloud completion accuracy and point cloud density.

The utility model discloses an inclined ladder gravity center transfer balance training plate which comprises a supporting part used for bearing, a treading part is arranged on the supporting part, the treading part is obliquely arranged on the supporting part and can swing back and forth relative to the supporting part, and a user stands on the treading part, so that the inclined ladder gravity center transfer balance training plate is formed. Slow and continuous traction can be achieved through the shaking and inclination angle of the treading part, the tension of the triceps calf muscle can be changed, ankle joint deformity can be corrected and prevented, the triceps calf muscle is drafted, foot drop is improved, deformity is prevented, and the extensibility and elasticity of the muscle are recovered. Two combined treading parts can be used, two feet can stand on the different treading parts, training and practicing can be carried out in cooperation with body twisting, and the fitness device is suitable for athletes or fitness enthusiasts and is low in use cost, small in occupied space and convenient to transfer.

This specification provides a training method, text processing method, and system for a large language model, comprising: obtaining a sample set, the sample set including multiple sample texts; obtaining a first large language model to be trained, the first large language model employing a linear attention mechanism; reusing at least some model parameters of the first large language model from pre-trained model parameters of a second large language model; the second large language model employing a non-linear attention mechanism; training the first large language model using the sample set; and obtaining a trained target large language model. This method can accelerate the training process, reduce computational and time costs, and improve the performance of the target large language model. Furthermore, it can optimize the computational complexity of attention to linear levels and optimize storage space.

Embodiments of the present application disclose a robot assembly training method, a robot assembly control method and a computer readable storage medium. The training method comprises: constructing an initial agent, generating a preset number of joint state combinations based on the initial agent; in a simulation environment, controlling a simulated robot to perform an assembly action according to the joint state combinations; determining a reward mode, scoring the performed assembly action based on the reward mode to obtain a reward score; and training the initial agent using the reward score. The control method comprises: obtaining a trained agent; inputting environment data of a current time into the agent to adjust the agent, and obtaining an implemented agent; and inputting a task instruction into the implemented agent, and performing an assembly action according to a strategy output by the implemented agent. Therefore, the present application can effectively improve the adaptability of the assembly robot and accurately complete the assembly action in more application scenarios.

This utility model discloses a lower limb rehabilitation training device, including a frame, a leg mechanism, and protective components. The leg mechanism comprises two identical sets, located on opposite sides of the frame. Each leg mechanism includes a drive rod, a third link, a fourth link, a first connector, a second connector, a pedal, and a driving component. A rotating shaft is rotatably mounted on one end of the drive rod, and the rotating shaft is rotatably mounted on the frame. The driving component controls the drive rod to rotate around the axis of the rotating shaft, which in turn drives the hinge of the third and fourth links to rotate, thereby controlling the movement of the first and second connectors. In conjunction with the fifth and sixth links, the pedal moves along a trajectory that most closely resembles the natural human gait, thus training the trainee to complete the natural human gait trajectory and improving the effectiveness of rehabilitation training. Furthermore, this application employs a multi-link structure, which provides stronger support, increases the load capacity and safety of the device, and, together with the protective components, provides comprehensive protection for the trainee.

The application discloses a coal mine image segmentation model and method based on a VMamba and multi-expert hybrid network and a construction method thereof. The coal mine image segmentation model is constructed. The image block encoding layer output of an encoder is taken as the input of the first encoding layer of a first VSS network, and the output of the first encoding layer of the first VSS network is taken as the input of the output end feature learning module. The input of the output end feature learning module is taken as the input of the first decoding layer of a decoder, and the output of the first decoding layer of the decoder is taken as the input of the second decoding layer of the decoder. The input of the second decoding layer of the decoder is taken as the input of a segmentation head, and the output of the segmentation head is taken as a segmented image. The application can significantly improve the segmentation precision and calculation efficiency of the coal image.

This invention discloses a deep learning-based method for ultra-short-term photovoltaic (PV) power prediction. First, historical data is preprocessed to eliminate irrelevant variables and accelerate model training. Then, the advantages of three clustering algorithms are combined to obtain a more reasonable dataset partitioning. Next, particle swarm optimization is used to optimize the parameters of variational mode decomposition. Finally, parallel learning of CNN and GRU networks is employed to identify local and temporal features of the data, enabling the network structure to fully leverage the input data. Finally, a deeper learning process is achieved through CNN neural network concatenation and fusion, resulting in high-precision prediction. The PV power prediction method of this invention demonstrates excellent performance, significantly outperforming other traditional models in predicting PV power under different weather conditions.

The invention relates to the technical field of medical information, in particular to a data supervision method and system for medical information management. The invention discloses a data supervision method for medical information management. The method comprises the following steps: S1, obtaining portable probe data and probe related bed data; s2, performing medical staff training judgment according to the virtual human body route use sequence and the actual human body route use sequence of the portable probe; s3, according to the probe type use sequence, calculating to obtain probe use entropies of the beds and sharing indexes among the beds; and S4, according to the difference points, the cross-personnel difference points and the to-be-detected bed sets, performing sequence position analysis on probe types in each to-be-detected bed set by using a sequence. According to the method, two core indexes including the probe use entropy and the bed sharing index are creatively provided, and the probe use entropy can sensitively capture the disorder degree of use of a single-bed probe.

This invention relates to a dual-modal flexible temperature-strain sensor and its fabrication method. The sensor achieves efficient detection of temperature and strain signals by combining direct ink writing and screen printing processes. The sensor features a unique design, comprising a rigid substrate and a flexible substrate, housing a thermal resistance temperature sensor and a piezoresistive strain sensor respectively, ensuring effective separation and accurate measurement of temperature and strain signals. The use of flexible materials and a FA-BP neural network algorithm based on an improved firefly algorithm enhances stability and measurement accuracy under varying environments. Furthermore, the sensor can still operate normally at high temperatures, outperforming common flexible thin-film materials on the market. This sensor demonstrates significant application potential in smart wearable devices, health monitoring, and industrial automation, providing a new solution for achieving high-precision and high-stability temperature and strain monitoring.

The application belongs to the technical field of neural network model training, and discloses a neural network model training method and system based on an improved random configuration algorithm, which comprises the following steps: under the condition of reducing the root mean square error of the neural network model output, screening suitable neurons as candidate hidden layer nodes through the inequality constraint condition of the random configuration algorithm; screening K neurons with the fastest training error reduction from the candidate hidden layer nodes, and selecting the neuron least related to the previous L-1 hidden layer nodes from the selected K neurons as the optimal hidden layer node; calculating the output weight through the least square method, and updating the structure of the random configuration network; and judging whether the network structure is completed by using the maximum allowable number of hidden layer nodes and the maximum allowable output error. The application can reduce the overall calculation amount of the algorithm while ensuring the prediction accuracy, and is suitable for application scenarios with high real-time requirements.

The application discloses a network algorithm integrated multi-core particle integrated system network design method, which designs the network layer in the multi-core particle integrated system by integrating an operation unit in a network unit. The method mainly comprises router micro-architecture design, task mapping and software and hardware collaborative design; the router micro-architecture design is the hardware implementation, flow control, pipeline design and arbitration design of the router, and part of the calculation amount that needs to be aggregated is transplanted from the processor to the network for processing; the task mapping is the mapping of the tasks in the application system flow diagram to the multi-core particle integrated system; and the software and hardware collaborative design is the optimal software and hardware configuration designed in the mode of software and hardware combination with the minimum application running time as the target. The application can realize calculation in the data transmission process, reduces the data amount of network transmission and communication delay, thereby accelerating the running speed of the application on the multi-core particle integrated system.

The utility model discloses a kind of multifunctional rehabilitation training equipment, including U-shaped walking aid support A and the U-shaped walking aid support B of parallel arrangement in U-shaped walking aid support A right side, the outer portion of U-shaped walking aid support A and U-shaped walking aid support B center place both ends is all sleeved and is fixed with metal fixed sleeve B, two groups the metal fixed sleeve B between are all connected with inner support connecting rod, two the inner support connecting rod center place front end upper side is provided with metal seat plate, by increasing the novel double-arm training device on the inner support connecting rod in U-shaped walking aid support A and U-shaped walking aid support B center place, when rehabilitation patient sits on metal seat plate and rests, can use double-arm training device to train double arm, especially elderly rehabilitation patient, only need to take out and hang in hanging pole back type handle, through back type handle ceaseless to and fro pull elastic training rubber band, to improve upper limb strength and joint flexibility, promote the overall rehabilitation of body.

The invention discloses a multi-modal physiological signal depression state recognition system based on data enhancement, and the system comprises a recognition module which is configured to input a multi-modal physiological signal sample into a trained depression state recognition model, and obtains a depression state recognition result; wherein the trained depression state recognition model is used for performing feature extraction on each preprocessed modal physiological signal to obtain a feature corresponding to each modal physiological signal; performing feature fusion on the features corresponding to the multi-modal physiological signals to obtain fused features; the fusion features are classified, and a depression state recognition result is obtained; wherein in the training process of the depression state recognition model, a training set is preprocessed, then data enhancement is performed on a preprocessing result, then the model is trained by adopting enhanced data, and when the total loss function value of the model does not decrease any more, training is stopped, and the trained depression state recognition model is obtained.

The invention discloses a honeycomb-free large-scale MIMO unmanned aerial vehicle associated power control method and equipment based on reinforcement learning, and a medium, and relates to the technical field of wireless communication. The method comprises the following steps: acquiring parameters of a communication system containing multiple unmanned aerial vehicles and access points; a cellular-free large-scale MIMO transmission model is constructed; the method comprises the following steps: determining a state space, an action space and a reward function based on a cellular-free large-scale MIMO transmission model, and respectively constructing intelligent agents for an uplink power control coefficient optimization problem of an unmanned aerial vehicle in a system and an access point unmanned aerial vehicle clustering optimization problem; training the intelligent agent by using a depth deterministic strategy gradient reinforcement learning algorithm; and obtaining an uplink power control coefficient of the unmanned aerial vehicle in the system and an unmanned aerial vehicle clustering result of each access point by using the intelligent agent based on an optimized uplink power distribution strategy and an access point unmanned aerial vehicle clustering strategy. According to the method, efficient joint optimization of unmanned aerial vehicle access point association and uplink power control can be realized under a rapidly changing channel condition, and the method has relatively high practical practicability under a low-altitude economic background.

The application relates to a training model for real-time pollution data prediction, which takes an LSTM as a basic model, accelerates the convergence speed of the LSTM through MAML, and reduces the limitation of the LSTM model data set. The application has the beneficial effects that the limitation of insufficient data volume of a pollution factor data set is broken through, and the balance between single factor prediction and multi-target prediction can be realized; the model can realize the function of accelerating the training of a prediction model while keeping the prediction accuracy unchanged; the model is light in volume, and the generalization of the training result on a small sample data set training is strong.

The application discloses to the technical field of artificial intelligence, computer vision and deep learning, and particularly relates to an image generation quality evaluation and optimization method based on sparse feedback for industrial design, which comprises the following steps: constructing an industrial design image dataset containing multi-dimensional expert scores; designing a physical flow branch to extract material texture and illumination statistical features, and a semantic flow branch to process picture-text splicing features through a self-attention mechanism; introducing a physical-visual outer product fusion module to capture second-order feature interaction; using the trained evaluation network as a reward model, combining a sparse sampling strategy and a LoRA fine-tuning mechanism, and guiding the diffusion model to align with the industrial design standard under extremely low computing resource consumption. The application significantly improves the material reality, illumination logic and picture-text consistency of the generated image, and provides an efficient and low-cost solution for automatic image generation in the field of industrial design.

The application discloses a kind of based on in-network aggregation's distributed model training task unloading method and device. Among them, the method is executed by the controller of data center network, the data center network includes several servers and programmable switch, including: according to the real-time network topology information in data center network, network resource information is constructed network physical topology model for each incoming distributed model training task;On the basis of the physical topology model, logical aggregation model is constructed by omitting the forwarding information of gradient data;According to the logical aggregation model and the available aggregation resources in the data center network, the task unloading strategy of distributed model training task is determined.The application is jointly optimized to worker, PS's assignment and in-network aggregation routing, to further accelerate model training, and reduce the communication overhead generated by training.

The invention discloses a communication method based on decentralized data-level federation forgetting, which comprises the following steps of: iteratively distributing undistributed data in a self-organizing network to each partition by adopting a heuristic partition method through a communication model of decentralized data-level federation forgetting based on a communication task, cooperative operation of the clients and neighbor clients in the self-organizing network through a communication topology formed by communication links is achieved, and communication distribution processing is completed. The self-organizing network comprises clients which are interconnected through a transmission link, and the clients comprise edge devices. The invention discloses a communication method and system based on decentralized data-level federal forgetting, and a storage medium, and the method comprises the steps: dividing a client into independent partitions which are not intersected with each other through a heuristic partitioning algorithm, and remarkably reducing the calculation and communication overhead; and a scoring strategy is introduced for guiding dynamic weighted fusion of the local model, so that the method has higher generalization ability and stability in a complex edge environment.

The application provides a steel structure strain data monitoring method based on a distributed optical fiber sensing technology, and belongs to the technical field of steel structure damage monitoring. A pre-play type layout design is performed based on a digital twin model, distributed optical fiber stress sensors are arranged at key positions of a steel structure, a partition backbone network architecture is constructed, the sensing network is divided into multiple monitoring sub-areas, each monitoring sub-area is connected to an edge computing gateway for data preprocessing and feature extraction, a hierarchical AI model is constructed based on obtained feature data, the hierarchical AI model is verified and optimized through virtual-real interaction between the hierarchical AI model and the digital twin model, damage positioning and type discrimination of the steel structure are performed by using the optimized hierarchical AI model, and steel structure strain data monitoring is completed.

The invention provides a human-like parking track planning method, system and device based on an attention mechanism and a medium, and relates to the field of automatic driving and reinforcement learning, and the method comprises the steps: S1, constructing a training environment for simulating a parking scene; s2, constructing an attention network by adopting a mixed attention mechanism according to training information in the parking scene; s3, training the reinforcement learning model, and adding a human preference-based guide mechanism into the model; s4, initializing an Actor network and a Critic network of the learning model, and training the learning model through the constructed simulated parking scene; and S5, based on the input training information in the parking scene, outputting a steering wheel rotation angle and acceleration instruction, obtaining an automatic parking track plan of the current automatic driving vehicle, and completing automatic parking. According to the invention, the efficiency and safety of automatic parking are improved, safer and more efficient automatic parking track planning is realized, and the driving track is more in line with human driving habits.

The utility model discloses a quick mounting mechanism for an aero seat ground rail, which relates to the related field of aero seat ground rails, and comprises a bottom plate, and a nut is welded at the lower end of the bottom plate; the fixing sleeve is welded to the upper end of the bottom plate; the clamping blocks are movably arranged at the side opening of the fixing sleeve; the pressing shaft is in sliding fit with the guide hole of the fixing sleeve, and the outer conical surface of the pressing shaft is in contact with the side conical surface of the clamping block; the ground rail is arranged on the periphery of the main body of the fixing sleeve in a sleeving manner; the bolt penetrates through the pressing shaft, the fixing sleeve and the bottom plate and is in threaded connection with the nut; the bolt is screwed down, the bolt drives the pressing shaft to move downwards in the axial direction, the outer conical face pushes the side conical face to enable the clamping block to move outwards in the radial direction, the ground rail is clamped, and therefore the ground rail and the fixing sleeve are connected. The installation process is simple and fast, the installation time is greatly shortened, and the installation efficiency of the aero seat ground rail is effectively improved.

The utility model relates to table tennis ball serving equipment manufacturing technical field, specifically speaking to a kind of structure is reasonable, rotates stably, and the action responds fast's table tennis ball serving machine nose rotating device, its characterized in that, the nose rotating mechanism is equipped with rotating motor (3), transmission gear, rotating worm (4), turbine (5), slip ring (6), wherein rotating motor (3) is fixed on rotating slipknot cover (2) by motor support, and rotating worm (4) is rotated by transmission gear, the turbine (5) is installed on ball channel outlet pipeline (1), and it is matched with rotating worm (4), and slip ring (6) is arranged between serving machine nose and rotating slipknot cover (2).

The present application relates to a kind of based on multi-element information fusion plug-in hybrid electric vehicle energy management method, comprising: obtaining adaptive fuzzy reasoning system model, vehicle speed sequence prediction model and SOC reference trajectory estimator;Real-time calculation current time transmission ratio, future finite time domain within the vehicle speed prediction sequence and future finite time domain within the SOC final value, based on Pontryagin minimum principle within the finite time domain rolling optimization, obtain the reference covariant variable within the time domain;If there is solution in rolling optimization, then use reference covariant variable in subsequent control domain to carry out open-loop Pontryagin minimum optimization, obtain corresponding control set, if there is no solution in rolling optimization, then execute SOC follow strategy.Compared with prior art, the present application will transmission ratio selection and torque distribution be jointly included in the category of energy management, and multi-element driving information is fused, and double-layer strategy of rolling optimization and SOC follow is proposed, which effectively improves the effect of plug-in hybrid electric vehicle energy management strategy.

The invention provides a communication method, terminal equipment and network equipment. The method comprises: a terminal device receives a first data set sent by a network device, the first data set including first data and not including second data, and the first data and the second data are both used for training, fine tuning or retraining of a model of the terminal device. In the application, the first data set sent by the network equipment to the terminal equipment can comprise the first data and does not comprise the second data acquired by the terminal equipment, and the terminal equipment can utilize the received first data in the first data set and the second data acquired by the terminal equipment to send the first data to the terminal equipment. According to the method, the model in the terminal equipment is trained, fine-tuned or retrained, and the second data acquired by the terminal equipment does not need to be transmitted at the air interface, so that air interface resources can be saved, the energy consumption can be reduced, and the acceleration of model training is facilitated.