130 results about "Training performance" patented technology

This application provides a method for sending training data of an AI model, which is applied to a scenario in which a reporting network element of the training data sends the training data to a training network element of the AI model. After obtaining the training data, the reporting network element classifies the training data into reference data and non-reference data, and indicates the training data by sending first-type information and second-type information to the training network element. The first-type information includes full information of the reference data, and the second-type information includes incremental information of the non-reference data relative to reference data corresponding to the non-reference data. Thus to help reduce overheads of sending the training data, thereby helping reduce air interface overheads in an AI model training or updating process, reduce a transmission delay, and improve model training performance.

According to the mechanical arm control method and system based on the self-adaptive force field, the motion state of a patient is monitored, the motion state is compared with a demonstration space trajectory, and training performance information of the patient is obtained. Based on the training performance information and the constructed adaptive force field model, auxiliary force applied to the patient is obtained through calculation, wherein the auxiliary force comprises normal force, tangential force and viscous force. And the applied auxiliary torque is calculated based on the tail end posture of the mechanical arm and the target posture, the joint control torque is obtained according to the auxiliary force and the auxiliary torque, and the dynamic torque and the null space control torque of the robot dynamic model are obtained. And the joint control torque, the dynamic torque and the null space control torque are combined to obtain the comprehensive control torque for controlling the mechanical arm. Therefore, the self-adaptive adjustment of the auxiliary force is realized based on the position, speed and direction of the patient, and the time degree of freedom can be obtained in the motion of the self-adaptive force field, so that the patient obtains more freedom of active motion under the assistance of the space force field.

The invention discloses an iterative supervision fine tuning data generation method and system, and relates to artificial intelligence and data processing technologies, and the method comprises the steps: generating a basic training data pair based on a pre-training language model, and the basic training data pair comprises an instruction-response pair; performing model training based on the basic training data, and determining training performance feedback data according to the performance of the model on the verification set and the task index; analyzing a model performance index according to the determined training performance feedback data; predicting and optimizing an optimization strategy of training data and training parameters based on the model performance index and the basic training data pair; and performing model training according to the optimized training data and the training parameters so as to execute AI model training platform data generation by using a strategy output by the model after training. Model performance feedback, automatic task mining and sample optimization mechanisms are introduced, training data are dynamically adjusted and expanded, and closed-loop iteration and performance-oriented optimization of data generation are achieved.

The invention discloses a DIKWP-driven individualized brain-map interactive feedback system, which is used for neural rehabilitation and brain-computer interface training. The system obtains brain activity data of a patient through electroencephalogram acquisition equipment, constructs a personal brain-semantic map in combination with cognitive evaluation, and establishes a mapping relation between semantic units and brain region responses. In the training process, the DIKWP semantic analysis module performs multi-layer analysis on indexes such as reaction time, accuracy and intention achievement, generates multi-mode instant feedback such as visual sense, auditory sense or tactile sense, and performs directional reinforcement on a weak semantic domain. The system has a dynamic target optimization capability, and the difficulty can be automatically adjusted according to training performance; when attention distraction, emotion abnormity or semantic deviation is detected, a safety intervention mechanism is automatically triggered, and training effectiveness and safety are guaranteed. According to the method, closed-loop individualized rehabilitation interaction is realized, the adaptation degree and efficiency of brain-computer interface training are remarkably improved, and the method is suitable for various rehabilitation scenes such as languages, movement and cognition and has a good industrial application prospect.

The invention discloses a method for establishing a rough-arrangement search model based on LTR, and the method comprises the following steps: S1, collecting candidate total data, and completing the preprocessing operation; s2, the sorting score of each piece of data is calculated and arranged in an ascending order, and the data are divided into equivalent data buckets; s3, randomly extracting data to form a training sample pair, generating a Pair-wise data set, and marking a sorting relationship; s4, executing rough-arrangement search model training and calculating a prediction score; s5, constructing a loss function by adopting a Lambda gradient signal, feeding back a sorting error and iteratively updating model parameters; s6, monitoring a training performance index, stopping training after a stopping condition is met, and deploying the model to an online environment; and S7, storing the trained model file, and defining an input format, a sorting structure and an output interface. According to the method, the sorting accuracy and fine sorting consistency of the rough sorting search model are improved, and the generalization ability of the model is remarkably enhanced.

The invention discloses a pulse neural network lightweight method based on nonlinear pruning, and relates to the technical field of artificial intelligence and neural networks. The method comprises the following steps: S1, constructing an NDI-LIF spiking neural network model, and introducing a bilinear product term into an input current in a dynamic updating process of a neuron membrane potential; s2, constructing a nonlinear synaptic pruning mechanism, representing a connection weight as a re-parameterization function composed of a re-parameterization weight and a conversion gain coefficient, and pruning the connection weight based on the re-parameterization function; and S3, training the NSPDI-SNN model to obtain a lightweight pulse neural network. By introducing a nonlinear dendritic integration mechanism and a state-adjustable synaptic pruning mechanism, balance between network sparsity and high performance is realized while the expression ability of the model is enhanced, and a lightweight spiking neural network model with high spatial-temporal expression ability, highly sparse structure and reasonable biological mechanism is constructed. And the generalization ability and the efficient training performance under various tasks are ensured.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as financial science and technology, medical health and the like, and discloses a federal training method, device, equipment and medium based on segmented sparse coding. Performing sparse processing in combination with historical residual errors, generating residual errors and sparse segments, and encoding and uploading the residual errors and sparse segments to a server; the server performs weighted aggregation on the segments with the same identifier according to the sample size and issues the segments; the client fuses global and local segments, updates low-rank adaptive parameters, stores residual errors and completes training; and after a preset round is completed, integrating the updated parameters and the pre-training model to form a fine tuning model, and processing task input data to generate a result. According to the method, the communication burden is reduced through segmented sparse compression and weighted aggregation, the adaptation effect is improved through exponential decay fusion, and the training performance and efficiency are both considered in a communication limited and privacy sensitive environment.

The invention relates to the technical field of intelligent robot training, and provides an intelligent robot cerebellum training system and method based on simulation and reality fusion, and the system comprises a simulation training terminal, a reality training terminal, a cerebellum learning simulation terminal, a data fusion and modeling terminal, a control execution terminal, and a central monitoring and evaluation terminal. The simulation training terminal is used for performing action training and generating an initial control strategy and simulation interaction data; the reality training terminal is used for executing an action task and collecting execution data; the data fusion and modeling terminal is used for receiving simulation interaction data and execution data and carrying out distribution alignment; the cerebellum learning simulation terminal is used for calculating a prediction error and outputting a correction signal; the control execution terminal is used for generating real-time state feedback; and the central monitoring and evaluation terminal is used for generating a global scheduling instruction according to the training data and the real-time state feedback. The method has the effects of reducing the training cost of the cerebellum of the intelligent robot with the body and improving the training performance.

The present application relates to the technical field of intelligent training, in particular to a kind of intelligent shooting examination training system and method, including the following contents: posture recognition module, trajectory monitoring module, hit analysis module, rhythm comparison module, label output module.In the present application, based on the motion image of the process of training personnel holding gun to firing, the synchronous angle trajectory of shoulder and elbow is identified in real time, compared with the combination of standard action, the matching difference of each period is accurately screened and quantified, the change of laser trajectory coordinate point and direction during aiming is analyzed, the action control stability is dynamically evaluated, the hit point sequence is used, the structured score is formed by combining path extensibility and hit area density, the start and end time of action unit and rhythm arrangement of training process are compared one by one, the phenomenon of advance and delay is identified, the performance score is flexibly adjusted according to the weight of multi-dimensional index, and the accurate matching between training performance evaluation and actual ability requirement is realized.

The invention discloses a visible light-infrared multi-mode target detection method based on transfer learning, and relates to the technical field of target detection. The objective of the invention is to solve the technical problem of poor model training effect caused by a small number of partial category samples in existing visible light-infrared multi-modal target detection. The method comprises the following steps: firstly, supplementing pictures for categories with insufficient sample quantity in a data set so as to expand small sample category data quantity; then, a weight obtained by pre-training on the single-mode data is used as an initial weight of the YOLOv11 model; and finally, the initial weight is loaded to a visible light-infrared multi-mode target detection training process, and model training is completed. According to the method, the performance of the YOLOv11 model in a multi-modal target detection task is effectively improved by multiplexing the single-modal pre-training weight through transfer learning and combining a small sample data supplement strategy, compared with a non-transfer learning detection scheme, the detection precision and stability are remarkably improved, and the method can be widely applied to scenes needing multi-modal target detection.

This invention belongs to the field of artificial intelligence hardware architecture design and discloses a method, system, device, and medium for cross-layer optimization of multi-core packaging and optical interconnect clusters for large language model training. The method includes: constructing a design space, including a core hardware architecture layer, an optical interconnect network layer, and a training parallel strategy layer; performing an outer-layer search to search for the architecture parameters of the multi-core modules within the core hardware architecture layer; for each outer-layer search sampling point, performing an inner-layer search to collaboratively optimize the optical interconnect network topology and training parallel strategy within the optical interconnect network layer and the training parallel strategy layer; and outputting the Pareto optimal design point on the training performance and cluster cost plane. The technical solution described in this invention can guide the design of related training clusters, thereby fully leveraging the advantages and potential of core technology and optical interconnect technology in large language model training.

The present disclosure relates to communication methods, communication devices, a communication system, a storage medium and a program product, and belongs to the technical field of communications. A method comprises: a first device receives second data corresponding to first data transmitted by a second device, wherein the first data and the second data form training data pairs, and the training data pairs are used for training a first model. In the method provided in the present disclosure, while transmitted data of a transmitting end is unknown, the transmitted data of the transmitting end and received data of a receiving end are combined as training data pairs on the basis of the received data, so as to train a model, thereby improving the model training performance and accuracy.

We provide the system. [Solution] A means for generating an individual training plan using a trainer generated based on a virtualized character selected by the user, A means of analyzing the user's exercise in real time and providing immediate feedback according to the training performance, A means of recording training results and managing rewards based on points awarded to users, A means of providing community features to share results among users and promote competition, A system that includes this.

The invention relates to the technical field of virtual experience, and discloses an immersive simulation training evaluation system and method based on virtual reality, and the system comprises an equipment interaction perception module which is used for capturing the sight line, gesture data and feature data of a user; the environment scene simulation module is used for reconstructing and rendering a three-dimensional model of a target scene and performing digital simulation; the task training and logic control module is used for establishing task process management, analyzing a user instruction and adjusting the difficulty degree of a task according to real-time training performance after the user instruction is executed; the evaluation analysis module is used for quantitatively scoring the training performance of the user and generating a personalized training scheme according to a scoring result; and the data management platform is used for uploading the data to the cloud for storage and sharing. According to the method, a whole-process closed loop from scene construction, task control to evaluation optimization is realized, and the system response efficiency is improved; and a personalized training scheme can be generated according to a scoring result, so that targeted training of the user is facilitated, and the training efficiency is improved.

The application provides a mechanical arm control method and system based on an adaptive force field, wherein the motion state of a patient is monitored, the motion state is compared with a demonstration space trajectory, and training performance information of the patient is obtained. Based on the training performance information and an adaptive force field model constructed, an auxiliary force, including a normal force, a tangential force and a viscous force, applied to the patient is calculated. Based on the end posture of the mechanical arm and a target posture, an auxiliary torque applied is calculated. According to the auxiliary force and the auxiliary torque, a joint control torque is obtained, and a dynamic torque of a robot dynamics model and a null space control torque are obtained. In combination with the joint control torque, the dynamic torque and the null space control torque, a comprehensive control torque for controlling the mechanical arm is obtained. In this way, adaptive adjustment of the auxiliary force is realized based on the position, speed and direction of the patient, time freedom is obtained in the motion of the adaptive force field, and the patient can obtain more active motion freedom under the assistance of the space force field.

The invention provides a sample set updating method and device and electronic equipment, and the method comprises the steps: carrying out the feature extraction of a training sample in a first sample set, obtaining a first feature of the training sample, and carrying out the space mapping of the first feature, and obtaining a second feature of the training sample; clustering the training samples to obtain a plurality of first training sample groups; for each training sample group, based on the second features of the training samples in the first training sample group, performing sample screening on the first training sample group to obtain a second training sample group corresponding to the first training sample group; and replacing each first training sample group in the first sample set with a corresponding second training sample group to obtain a second sample set. According to the invention, the training performance of the sample set for training the machine learning model can be effectively improved.

According to one exemplary embodiment, an apparatus, system, method and / or computer program product may provide an athletic training performance analysis apparatus including at least one athletic training performance analysis device including at least one user interface coupled to the at least one athletic training performance analysis device configured to interact with a user to receive a user selection of at least one training performance analysis routine, the at least one user interface may include: at least one electronic computer processor; at least one input device coupled to the at least one electronic computer processor to receive the user selection; at least one output device coupled to the at least one electronic computer processor to provide output to the user; at least one electronic memory coupled to the at least one electronic computer processor; and at least one of: wherein the at least one electronic computer processor is configured to save or retrieve the at least one training performance analysis routine from the at least one memory; where the at least one electronic computer processor is configured to electronically monitor and electronically analyze the at least one training performance analysis routine; wherein the at least one electronic computer processor is configured to randomize at least one challenge by the at least one athletic training performance analysis device; or wherein the electronic computer processor is configured to at least one of: combine a plurality of previously saved of the at least one training performance analysis routine, or shuffle a plurality of previously saved of the at least one training performance analysis routine.

The invention discloses a reinforced training method for improving the middle-section text analysis capability of a large language model, and relates to the field of natural language processing and generative large language models.The method comprises the steps that middle-section double-window construction is conducted according to a dynamic scale factor, extraction and splicing are conducted from original long text corpora based on window positions, and the middle-section text corpora are obtained; obtaining a preliminary training sample; performing middle text information density evaluation on the preliminary training sample, and verifying and optimizing the preliminary training sample to obtain an optimized sample; and inputting the optimized sample and the updated position identification sequence into a large language model, and realizing iterative training of the large language model through a position-aware middle loss weighting mechanism to obtain an optimized large language model. According to the method, the retrieval and understanding accuracy of the model in the process of processing the middle position information of the sequence can be remarkably improved in a balanced manner, the U-shaped performance curve which is sunken originally is successfully leveled, and the training performance bottleneck is overcome.

Provided is a load control unit with which a user of a training instrument can immediately ascertain his / her performance and optimize training. A load control unit (10) is installed in a training instrument (1) and controls the load of the training instrument (1), the load control unit (10) executing first display processing for causing a display device (30) to display the target speed of at least a part of one stroke relative to the stroke amount of a user who uses the training instrument (1), detection processing for detecting the speed of the stroke of the user and the stroke amount of the user for each prescribed time, and second display processing for causing the display device (30) to sequentially display the speed with respect to the stroke amount for each prescribed time in accordance with the target speed.

The communication method and device provided by the embodiment of the invention are used for improving the training performance. The method comprises the following steps: receiving first capability information from first management equipment, wherein the first capability information indicates training capability supported by the first management equipment and / or network element equipment; and sending a training request to the first management device according to the first capability information, wherein the training request is used for requesting to execute a first reasoning function by adopting a first training technology. According to the mode, the second management equipment can obtain the training technology supported by the first management equipment and / or the network element equipment, so that the second management equipment can select the appropriate training technology for the specific reasoning function according to the requirement to trigger the first management equipment and / or the network element equipment to execute training.

The invention provides an agent communication strategy optimization method based on reinforcement learning, and the method enables a system to adaptively select an optimal beam forming and interference strategy according to a real-time channel state through introducing a reinforcement learning technology and combining the performance evaluation of a pre-training model, thereby achieving the optimization of an intelligent agent communication strategy. The balance between the maximization of the communication quality of the target user and the most intensified interference effect of the non-target user is realized, and the reliability and safety of the system are guaranteed in a complex dynamic environment. By constructing the pre-training performance mapping table and guiding the reinforcement learning algorithm to adaptively select the optimal model strategy according to the channel state, the receiving capability of the unauthorized user is effectively interfered while the communication quality of the target user is guaranteed, and the system security is improved. According to the method, efficient, self-adaptive and intelligent communication interference cooperative control can be realized, and the robustness, the anti-interference capability and the transmission quality of the system are effectively improved.

The invention provides a model training method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence system performance optimization and deep learning framework scheduling. The method comprises the following steps: in a current batch training process of a target model, acquiring execution information of a corresponding bottom kernel when a processor executes a plurality of computational operators of the target model; according to the execution information, determining a global key path, related to the total calculation duration, of the target model in the current batch training process, and identifying target operators forming training iteration delay in calculation operators contained in the global key path; and generating a delay optimization instruction for the target operator, and in response to the delay optimization instruction, adjusting training configuration parameters corresponding to the target operator in an upper-layer training framework of the processor so as to apply the adjusted training configuration parameters in a subsequent batch training process of the target model. According to the scheme, the training performance of model training can be improved.

The invention provides a distributed training method and device of a large model, electronic equipment and a storage medium, and relates to the technical field of model training. The invention provides a gradient compression and expert network parallel collaborative cross-domain distributed training method. In the training process, gradient synchronization and expert parameter routing between the target expert sub-networks are controlled through a proposed collaborative compression and dynamic scheduling mechanism, the cross-domain communication traffic can be remarkably reduced, and the situation that the training efficiency is reduced due to insufficient network bandwidth is avoided. Meanwhile, on the premise that the model precision is kept stable, adaptive adjustment of the expert routing strategy and the gradient compression mode can be achieved, and the distributed training performance of the to-be-trained large model under the multi-cluster and heterogeneous computing power condition is improved. The method can effectively support a large-scale model training task in a cross-data center, cloud edge cooperation and computing power network environment, and can be called by a training platform, a scheduling system and a computing power service system as a bottom layer capability.

The invention provides a data processing method and a related device, which are used for a first device to perform data enhancement on first training data so as to improve the quantity and diversity of the training data. In this way, training of a transceiver model (such as a transceiver neural network) is facilitated, and the training speed and training performance of the transceiver model are improved. The method comprises the steps that a first device receives first training data from a second device; and the first device performs data enhancement on the first training data to obtain enhanced training data.

This invention discloses a load-balanced automatic parallel modeling method, device, and storage medium. First, it analyzes the key factors affecting the execution performance of operators and models (operator in-degree, tensor shape, and tensor data type), and proposes a method for constructing a performance evaluation model based on operator characteristics to assess the computational, communication, and overall costs of operators, as well as the training performance costs of the model. Then, with the goal of load balancing across devices, a layer-by-layer partitioning method based on topological sorting is used to quickly divide the neural network model into multiple sub-models with balanced overall costs, achieving coarse-grained partitioning. Finally, based on the model's training performance evaluation model, and according to the communication characteristics between operators, a fine-grained model partitioning and scheduling method based on communication optimization is used to fine-grainedly adjust the coarse-grained sub-models, reducing the amount of cross-device communication tensor transmission to achieve optimal global model scheduling.

This invention discloses a knowledge graph embedding method based on negative sampling and weighting. The steps include: acquiring a medical dataset; sampling high-quality candidate negative sample sets for all positive triples based on pre-trained embedding vectors and a designed high-quality candidate negative sample set construction strategy; calculating the scoring function values ​​of triples and their corresponding negative triples according to a translation-based knowledge graph embedding model; statistically analyzing the relevant frequency data in the dataset and calculating the frequency-based weights for each triple; calculating the loss L of the triple set based on the scoring function values ​​and weights to obtain the final loss function of the training framework; optimizing the loss function of the training model using stochastic gradient descent; and outputting the potential association between drugs and diseases / symptoms. This application improves the negative sampling quality of the model and mitigates the adverse effects of long-tail entities and relationships on model training by weighting the loss function, thereby improving model training performance.

Provided is a method for training a quantized CNN by applying trend-based progressive scaling optimized for hardware accelerators. The method for training a quantized CNN according to an embodiment of the present invention determines a next suitable scale by identifying a previous trend, and trains the quantized CNN by applying a scale factor considering clipping loss sensitivity and progressively adjusting the scale factor on the basis of previous overflow / underflow. Accordingly, training based on hardware accelerators can be performed quickly and with minimal performance degradation, and training performance can be improved by facilitating higher precision utilization than before.

The invention relates to the field of three-dimensional reconstruction, and discloses a high-performance Gaussian splash training method and system based on a mercuric chloride intelligent chip, and the method comprises the steps: carrying out the parallel execution of streaming preprocessing and feedback scheduling on a host, and carrying out the incremental three-dimensional reconstruction to generate a new three-dimensional point and carry out the pose correction; meanwhile, asynchronous training and cooperative control are executed in parallel on the mercuric chloride intelligent chip, and NPU rendering and gradient back propagation are carried out; the core lies in that a set of closed-loop feedback mechanism from equipment to a host is constructed; the equipment analyzes the basic type of an error through differential gradient attribution, and feeds gradient attribution information back to the host through an asynchronous queue; and the host executes a heterogeneous scheduling strategy according to the information to realize intelligent cooperation of reconstruction and rendering tasks. The invention further provides a corresponding system. According to the scheme, front and rear end tasks are decoupled, the training performance and the hardware utilization rate are remarkably improved, convergence of a high-quality model is accelerated through intelligent feedback, and the stability of the training process is guaranteed.

Systems and methods are provided for a machine learning (ML) pipeline with a unified framework. A ML pipeline trains a ML model in a development environment, and further executes the ML model in a production environment. A data storage in the development environment stores training performance metrics corresponding to the training of the machine learning model. A development web server in the development environment retrieves the training performance metrics from the data storage in the development environment, and presents the training performance metrics. A data storage in the production environment stores production performance metrics corresponding to the executing the ML model in the production environment. A production web server in the production environment retrieves the production performance metrics from the data storage in the production environment, and presents the production performance metrics.

The invention relates to the technical field of intelligent health management, in particular to a pelvic floor muscle training system based on a mobile terminal, which comprises a signal collaborative acquisition module, a multi-parameter linkage recognition module, a three-section action decomposition module, a time sequence voice guide module and an execution deviation evaluation module. According to the method, through synchronous time sequence analysis of pelvic floor electromyographic signals and heart rate signals in the training process, collaborative changes of signal mutation nodes are mined, multi-source physiological parameter linkage recognition and induction of action stage morphological characteristics are combined, training actions are disassembled into multi-stage processes, and time sequences of signal nodes and voice guidance are dynamically mapped; precise butt joint of step-by-step voice instructions and action states is achieved, performance of each stage of the training process is quantified into data features, training performance differences are visually evaluated through the time sequence deviation degree, scientificity and continuity of rehabilitation guidance are enhanced through a step-by-step feedback mechanism, and data basic support is provided for individual rehabilitation path adjustment and progress tracking.