319results about How to "Improve training efficiency" patented technology

The invention discloses a photovoltaic power prediction method based on improved empirical mode decomposition and an optimized long short-term memory network, and the method comprises the steps: firstly carrying out the preprocessing of abnormal value elimination, missing value filling, normalization and the like of photovoltaic power and related meteorological data, and improving the data quality; then, an improved empirical mode decomposition (EE-ANEMD) algorithm is adopted to decompose the preprocessed power sequence into a multi-scale intrinsic mode function component and a residual term, and high-frequency noise, intermediate-frequency fluctuation and a low-frequency trend are effectively separated; global optimization is carried out on the hidden layer unit number, the initial learning rate and the maximum number of training times of the LSTM network through an improved sparrow search algorithm (ISSA), finally, the optimized LSTM is utilized to carry out training prediction on each component, and results are fused and subjected to reverse normalization to obtain a final value. Experiments show that the test set RMSE of the method is reduced compared with that of a single LSTM, the mid-term prediction precision is remarkably improved, and reliable technical support is provided for power system dispatching, new energy consumption planning and photovoltaic power station operation and maintenance.

The embodiment of the invention is suitable for the technical field of artificial intelligence, and provides a re-calculation training strategy generation method, electronic equipment and a computer program.The method is applied to a heterogeneous cluster and comprises the steps that cluster information of the heterogeneous cluster and model feature information of a to-be-trained model are determined; for each storage and calculation resource in the heterogeneous cluster, according to the cluster information and the model feature information, re-calculation time information and re-calculation performance requirements are determined; determining a re-calculation time information characteristic value under the condition that the re-calculation performance demand is not greater than the performance limit value of the storage and calculation resources; and determining a recalculation training strategy for the to-be-trained model according to the recalculation time information feature value corresponding to each storage and calculation resource. According to the embodiment of the invention, the adaptive re-calculation training strategy can be determined for each storage and calculation resource of the heterogeneous cluster, and the model training efficiency and throughput are improved by adopting the re-calculation training strategy corresponding to each storage and calculation resource to carry out model training.

The invention discloses a high-density electrical method infiltration surface three-dimensional inversion method based on multi-source data collaboration, and the method comprises the steps: constructing a multi-source heterogeneous data collection network, synchronously obtaining four types of data of a dam body monitoring region, and achieving the unification of coordinates and time scales of the four types of data through a space-time registration algorithm; establishing a multi-scale coupling model; on this basis, a deep learning inversion framework fused with physical constraints is developed; an inversion result is optimized through a self-adaptive dynamic correction module; and outputting the corrected three-dimensional distribution model of the infiltration surface. According to the method, a resistivity-moisture content-pore structure-fluid pressure quaternary coupling equation is macroscopically established through a multi-scale coupling model, pressure data are obtained in real time in combination with optical fiber sensing, pressure and pore deformation are correlated through a mesomechanical equation, a physical mechanism is fully fused, resistivity deviation caused by pressure is eliminated, and the resistivity is measured. The three-dimensional form of the infiltration surface is more suitable for the real situation, and an accurate position basis is provided for dam body permeability stability evaluation.

The invention discloses a deep face forgery detection model training method and device based on reinforcement learning, and the method comprises the steps: firstly obtaining a training set, and constructing a combined state vector containing current state information and historical state information for a sample in the training set; inputting the combined state vector into a tutor agent to output a loss weight; the method comprises the following steps: calculating a loss weight of a student detection model, weighting the original loss of the student detection model according to the loss weight, updating the parameters of the student detection model by using the weighted loss, calculating a reward signal according to the performance change of the student detection model before and after the parameter updating, and updating the strategy of a tutor agent by using the reward signal. The student model is dynamically guided to pay attention to difficulty and key samples through the tutor agent, the robustness and generalization ability of the deep face forgery detection model can be remarkably improved, and therefore a more reliable detection means is provided for deep face forgery.

The federal knowledge distillation method and application for rubber tire production parameter sharing divide the federal learning process into preheating and distillation two parts, the preheating part adopts the FedAvg federal training strategy, the distillation part utilizes the knowledge generated by each client, carries out the binned formation grouping knowledge, carries out the aggregation after the distribution to each client for the knowledge distillation in the server side.The invention proposes a new federal knowledge distillation method, through the method in the invention, the distributed training of the production line prediction model can be realized, and the model prediction accuracy of the server side is also higher than that of ordinary training.The invention can be used for the distributed training of the rubber tire production line key parameter prediction model, to improve the training efficiency, and in the case of not sharing data, the training result generated by the data can be shared, and the privacy can be well protected.

This application relates to the field of network communication technology, and in particular to a path pre-planning method, apparatus, and related equipment. The method includes: statistically analyzing the traffic presence of each training stream in each time slice of the current iteration cycle, wherein an iteration cycle is divided into M time slices, each time slice having the same duration; determining whether there is any overlap in traffic transmission time between the training streams based on the traffic presence of each training stream in each time slice of the current iteration cycle; and pre-planning the forwarding path of each training stream in the parameter network in the next iteration cycle based on the overlap in traffic transmission time between the training streams and the device and interface information of each training stream flowing through the parameter network.

A robustness measurement method for LeNet-5 networks based on adversarial boundary constraints is presented, relating to the field of deep learning model testing. The main steps include: for each training sample, dynamically generating adversarial examples based on the model's current state during training iterations; constructing a composite loss function based on standard cross-entropy loss and dynamic boundary constraint loss; performing end-to-end training on all parameters of the LeNet-5 network; and using the overall approximate robustness boundary as a measure of model robustness after training. This method effectively improves the resistance of the LeNet-5 model to fast gradient sign-based adversarial attacks without altering the basic structure of the LeNet-5 network by designing a new loss function.

The application provides an acoustic model training method and device. The acoustic model training method provided by the application comprises: acquiring unannotated acoustic samples; pre-training the acoustic model based on a self-supervised method, the acoustic model being used to predict the category of an input acoustic signal, in the self-supervised method, respectively generating masks for the time domain feature and the frequency domain feature of the acoustic signal based on the time domain feature and the frequency domain feature of the acoustic signal, the mask position and the mask quantity of different samples being different; dividing the levels to which the network structures of the pre-trained acoustic model belong, respectively freezing the layers of different levels, and asynchronously fine-tuning the acoustic model, the parameter freezing time of the layers of different levels being different, and the learning rate of the layers of different levels being different. The acoustic model training method and device provided by the application not only reduce the dependence on large-scale artificial annotation data, but also improve the training efficiency and task performance, and better adapt to task requirements and complex application scenarios.

The material detection image defect enhancement method based on multi-light source complementation of the application comprises the following steps: collecting and identifying parameters, processing the parameters by a ternary mapping model, and outputting noise feature prediction values and multi-light source initial parameters; adjusting multiple light sources after optimizing the multi-light source initial parameters to obtain optimized parameters, and collecting a noisy image of a material to be detected through a detection camera; inputting the noise feature prediction values, the optimized parameters, and the noisy image into a preset conditional GAN anti-noise model with light source constraints, and generating a defect enhancement image; when the parameter optimization criterion is not met, adjusting the optimized parameters, collecting a new noisy image, and repeating the steps until the parameter optimization criterion is met, and outputting the defect enhancement image, through the cooperation of multi-light source parameter adjustment and the introduction of the conditional GAN, a defect enhancement image with obvious and accurate features is generated, and through the introduction of an artificial intelligence function library and computer visual and auditory training and reinforcement of the model, the accuracy of subsequent defect recognition can be greatly improved.

The invention discloses a variable length sequence-oriented model training method, which comprises the steps of determining a training normal form, a training sample and a micro-batch size corresponding to model fine tuning according to a fine tuning request for a target model; the training sample is a variable-length sequence; dividing the training samples into micro-batches according to the sizes of the micro-batches so as to obtain micro-batch samples, and inputting the micro-batch samples into an assembly line stage so as to convert the micro-batch samples into intermediate representation; performing shape adjustment on the intermediate representation according to the sequence length of the training sample, the micro-batch size and the role type of the assembly line stage when the intermediate representation is transmitted between the adjacent assembly line stages; in the attention calculation process in each assembly line stage, performing attention weight constraint on the intermediate representation input into the stage by using an attention mask; in the pipeline stage, shape adjustment and attention weight constraint are executed through forward propagation, back propagation is completed to calculate a gradient, and parameters of a target model are updated. According to the invention, the training efficiency is improved and the computing resource consumption is reduced.

The application provides a training method and device of an image enhancement model, electronic equipment and a storage medium, and relates to the technical field of image processing. The method comprises the following steps: generating a low-illumination sample image corresponding to each normal-illumination sample image in an initial image data set according to the normal-illumination sample image, and generating a first training data set according to each normal-illumination sample image and the low-illumination sample image corresponding to the normal-illumination sample image; performing image mask extraction according to a target object in each normal-illumination sample image to obtain target mask information corresponding to each normal-illumination sample image; and training an image enhancement model according to the first training data set and the target mask information corresponding to each normal-illumination sample image. The method can improve the training efficiency and generalization performance of the image enhancement model and improve the enhancement effect of the image enhancement model on low-illumination images.

The present application relates to the technical field of short-circuit current calculation, and particularly relates to a flexible DC system short-circuit current calculation method based on a physical information neural network, comprising: setting a model input feature vector, the model input feature vector being used to represent a system operating state before a fault and fault information, and performing data preprocessing on the model input feature vector; constructing a hybrid driving calculation model, the hybrid driving calculation model comprising a physical calculation module and a neural network module, and being coupled based on a preset fusion architecture; performing end-to-end training and optimization on the hybrid driving calculation model by using a preset composite loss function; and performing flexible DC system short-circuit current calculation based on the optimized hybrid driving calculation model, so that the problems of poor precision, speed and convergence, poor interpretability and weak generalization ability in the prior art are solved.

The invention provides a data processing method and device for instrument detection, equipment and a storage medium, relates to the technical field of digital instrument detection, and solves the problem of insufficient detection precision caused by the technical bottlenecks of data scarcity and difficulty in small target recognition in a power distribution room scene in the prior art. According to the scheme, a large amount of sample data can be automatically generated based on a programmed image generation mechanism, and the target detection model introduced with the convolution attention mechanism is trained, so that a more accurate model is deployed on detection equipment; therefore, high-precision and high-robustness detection of the small-size and low-contrast digital instrument in the to-be-detected place under the complex environment of the power distribution room is realized, and the detection precision of the characters in the digital instrument is effectively improved.

This invention relates to the field of space mission planning technology, and provides a space multi-objective mission planning method that integrates reinforcement learning and curriculum learning. The method includes: constructing a space multi-objective mission simulation environment comprising an orbital dynamics model, an objective characteristic model, and a spacecraft constraint model; designing a progressive task sequence for agent training based on a curriculum learning method; and training the agent in the simulation environment using the progressive task sequence based on a deep reinforcement learning framework, updating network parameters by collecting experience data, and finally obtaining the optimal sequence decision-making strategy. This invention solves the technical problems of low decision quality, poor training efficiency, and weak policy adaptability caused by complex environments and sparse rewards in space multi-objective mission planning by guiding reinforcement learning through curriculum learning for stable and efficient training.

The invention belongs to the technical field of artificial intelligence, and particularly relates to a medical image data processing method and device and electronic equipment. The method comprises the following steps: performing feature extraction on medical image data to be processed to obtain image embedding features of the medical image data; according to a pre-trained image strategy model, model reasoning is carried out on the image embedded features, a candidate strategy set is obtained, the candidate strategy set comprises a plurality of candidate operation strategies, and the candidate operation strategies are used for representing operation rules for processing the medical image data; sending the candidate strategy set to a plurality of expert clients, wherein the expert clients are used for performing advantage and disadvantage evaluation on the candidate operation strategies according to local data to obtain strategy scores; and updating model parameters of the image strategy model according to the strategy scores returned by the plurality of expert clients. The processing efficiency and accuracy of the medical image data can be improved.

The application discloses a feature placement method and system for multi-GPU sampling graph neural network training, uniformly models a sampling graph neural network training process, and obtains hardware topology information; in a pre-sampling process, the number of times of accessing node features is counted as node heat, the node features are sorted according to the node heat, the node features are divided into a plurality of feature blocks according to a single-block capacity, and the feature blocks are divided into hot blocks or cold blocks according to block heat; based on the feature block set obtained by the division and the obtained hardware topology information, a mixed integer programming model is constructed, with the minimum bottleneck link time in a feature extraction stage and the memory load imbalance degree as targets, the mixed integer programming model is solved, and the placement position, access path and link flow of the globally jointly optimized hot block are obtained. The application can obtain a high-quality feature placement scheme under a given hardware constraint, minimize the extraction time, and improve the resource utilization rate of a multi-GPU system.

The application discloses a kind of battery core temperature online estimation and prediction method, system, equipment and storage medium based on mixed TSDM-AMBO-GRU, applied to lithium ion battery core temperature estimation and prediction field, comprising: obtaining battery operating data, and utilizing TSDM to create synthesis of battery operating data, as training set with battery operating data together;Training set is input to BO-GRU-AM framework, and the battery core temperature prediction model based on BO-GRU-AM is obtained by training;Input test set to battery core temperature prediction model, and obtain battery core temperature prediction result.The application not only can effectively overcome the problem caused by uneven distribution of temperature sensors and limited data in lithium ion battery system, and can realize real-time, efficient and accurate battery core temperature prediction under various temperatures and complex conditions, provide reliable fault warning and real-time monitoring support for BMS in practical application.

The application relates to the field of high-performance computing, and provides an HPC job power consumption prediction method and system based on power consumption curves and script information. The method comprises the following steps: obtaining a first similarity value based on historical job power consumption curve data; obtaining a second similarity value based on historical job script information data; assigning weights to the first similarity value and the second similarity value according to requirements, calculating a weighted sum, obtaining a comprehensive similarity value, and constructing a similarity adjacency matrix in this way; dividing HPC jobs into different categories according to the similarity adjacency matrix on the principle of maximizing a module degree index; training different neural network models based on historical data in different categories after the division, obtaining trained neural network models; matching a historical job category similar to a target HPC job, and using the neural network model of the historical job category to predict script information data of the target HPC job to obtain a prediction result.

This utility model relates to a training system and hand probe device for security personnel, comprising: a training uniform, a mannequin, and a hand probe device; the training uniform is worn on the mannequin, and a wireless communication tag is affixed to a predetermined position on the inside of the uniform; the hand probe device communicates wirelessly with a client, and one or more wireless communication tag readers are installed within the hand probe device. This utility model facilitates in-depth analysis and accurate feedback of training content, effectively avoids the subjectivity and arbitrariness of manual assessment, ensures the objectivity and consistency of assessment results, and helps improve the quality and effectiveness of security personnel training.

This invention provides a clustered federated multi-task learning method and apparatus for the Internet of Things (IoT). By clustering IoT terminal devices, the data distribution within the same cluster becomes more approximate. A federated multi-task learning algorithm is executed within each cluster, with global training and personalized training tasks performed on each IoT terminal device. This achieves data sharing within the cluster while fully utilizing local data from each IoT terminal device for training on personalized tasks, thus efficiently utilizing local data and improving training effectiveness. During local training on each IoT terminal device, the number of training rounds is adjusted based on computing power, fully utilizing the computing resources of each IoT terminal device and improving model training efficiency. Regularization constraints applied to personalized training tasks using the global model effectively prevent overfitting, control the degree of personalization, and improve model quality.

The invention relates to a circuit board defect detection method, device and equipment based on multiple sub-models and a storage medium, and is applied to the field of defect detection, and the method comprises the steps: obtaining an original image of a to-be-detected circuit board, aligning the original image according to a preset mother layout, and setting the aligned original image as a to-be-detected panoramic image; segmenting the panoramic image to be detected into a plurality of unit detection images according to a preset segmentation standard; matching a special sub-model from a preset detection model library according to the unit detection image, and performing defect detection on the unit detection image by adopting the special sub-model; and integrating the defect detection results of the unit detection images, and outputting the defect detection result of the circuit board to be detected. The circuit board defect detection method has the technical effect of improving the efficiency and accuracy of circuit board defect detection.

The application belongs to the field of smart grid and computer vision, and particularly relates to a power station wiring diagram text robust generalization detection and recognition method based on an improved SwinTextSpotter v2. The method comprises the following steps: step 1: inputting an image into a text detection and recognition network based on multi-modal learning for training and prediction, obtaining a shared feature map through a shared feature extraction backbone network, and further inputting the shared feature map into a text detection module to obtain a text detection result and a text feature map; step 2: inputting the text feature map into a visual feature extraction and prediction module to obtain a feature sequence, and then matching the predicted feature sequence with a canonical representation obtained by a character structure feature extraction and prediction module to obtain a recognition result; and the like. The application robustly improves the detection and recognition accuracy of the model for irregular text and Chinese character text, and improves the generalization performance of the text detection and recognition of various types of wiring diagrams.

The invention discloses a sparse time sequence Bayesian network construction method and system based on power distribution network topological constraints, and the method specifically comprises the steps: constructing a power distribution network topological graph, and calculating an adjacent matrix between nodes and a k-hop neighborhood matrix Nk; based on the power distribution network topology distance information and the matrix Nk, generating a hard constraint rule, and constructing a topology dependence mask matrix M; calculating an electrical influence range of the fault point on surrounding nodes to obtain an electrical influence matrix E; if the electrical influence coefficient of one node on the other node is smaller than a set value, deleting the corresponding dependent edge; introducing a data source reliability matrix R, and performing hard deletion or soft weakening on edges with reliability lower than a threshold value; combining the matrixes M, E and R to synthesize a sparse structure matrix S; and taking the matrix S as a space skeleton, adding a time dimension autoregression edge, and constructing a complete sparse time sequence Bayesian network. According to the invention, by guiding the rarefaction of the network structure, the number of network edges and the number of parameters are effectively reduced, and the trainability and reasoning efficiency of the model are improved.

A lithium battery data reconstruction and classification method driven by adversarial learning and feature distribution alignment, relating to the field of lithium battery state monitoring technology, mainly includes the following steps: constructing a feature recognizer to capture the temporal correlation features of lithium battery data throughout its entire life cycle; integrating the feature recognizer into an adversarial training framework, aligning the features of the data generated by the feature extractor with those of the real data through adversarial learning between real lithium battery data and simulated data generated by the feature extractor; constructing a feature transfer-based data reconstruction network based on the feature extractor and combining it with the Unet network to reconstruct battery data consistent with the features of the real data, forming an expanded dataset with the original data; and building a battery state classification network based on the feature recognizer, training it with the expanded dataset to complete the state classification of lithium batteries. This method ensures the effectiveness of the reconstructed data through feature distribution alignment, significantly improving the generalization ability and recognition accuracy of the classification model.

The application provides an X-band broadband high-gain radar metasurface antenna and a design method thereof. The metasurface antenna comprises, from top to bottom, a metasurface layer, an intermediate metal layer and a lower microstrip feed line layer. The metasurface layer comprises at least one metasurface unit. An asymmetric chamfer structure is designed on each metasurface unit. A center fork-shaped slot is arranged at the center of each metasurface unit, and four embedded parasitic branches are arranged at four side edges, respectively. Rectangular coupling slots corresponding to each metasurface unit are arranged on the intermediate metal layer. Feed lines corresponding to each metasurface unit are arranged on the lower microstrip feed line layer. The design method provided in the application can improve the design efficiency.

The application discloses a kind of super surface regulation and control device and method based on physical constraint generation model, it is related to electromagnetic super surface scattering regulation and control technical field.Method includes obtaining super surface array and forms different coding array structure, i.e.RCS response;Obtain simulation dataset and calculation dataset, establish including prediction model, identification model and generation model physical constraint depth condition generation model, utilize calculation dataset and simulation dataset to construct formula-simulation residual error network expansion dataset, pre-training is carried out to prediction model, and simulation dataset is used to the joint training of prediction model, identification model and generation model;Input target RCS condition, output optimal super surface unit arrangement scheme, realize the quick simulation and reconstruction of RCS.The application can reduce electromagnetic simulation workload, solve the problem that one-to-many mapping in traditional neural network is difficult to accurately train, and improve the intelligence and stability of super surface simulation RCS in combination with mechanical regulation.

The invention discloses a spatial orientation real-time analysis method based on eye movement, relates to the related technical field of flight training, and aims to overcome the defects of low analysis accuracy and poor adaptability in the prior art. The method comprises the following steps: acquiring pilot eye movement data through eye movement tracking equipment, acquiring head space orientation data through a flight helmet space sensor, and acquiring environment data containing instrument visual complexity and aircraft attitude from a virtual flight simulator; performing time synchronization on the three types of data; extracting features such as coordinates of a fixation point from the synchronous eye movement data; comparing the head orientation data with the reference data to generate deviation data; inputting the eye movement characteristics, the deviation data and the environment data into an evaluation model; the model dynamically adjusts parameters according to environment data, and calculates pilot space orientation capability indexes in combination with eye movement characteristics and deviation data. According to the method, accurate integration of multi-dimensional data can be realized, the analysis comprehensiveness and reliability are improved, a dynamic flight scene is adapted, and real-time accurate evaluation is realized.

The application provides a three-dimensional part retrieval method based on graph similarity search, relates to the field of computer graphics, and solves the technical problem that the prior art cannot fully utilize the geometry and design semantic information of CAD, and it is difficult to capture and display the topological relationship, resulting in low retrieval efficiency. The method comprises the following steps: obtaining part three-dimensional data of a computer-aided design (CAD) model, and constructing a training data set; constructing an edge-face connection graph based on the part three-dimensional data; calculating a graph edit distance (GED) matrix of all edge-face connection graphs in the training data set as a supervision signal; training a ranking model based on the GED matrix, mapping the edge-face connection graph to a hidden space, and constructing a part vector database according to the output graph-level embedding vector; the ranking model is constructed based on a graph attention network; inputting a CAD part to be queried into the trained ranking model to obtain a feature vector, performing nearest neighbor search in the part vector database, and returning a similar part result.

The application discloses a data processing method, device and equipment and a computer readable storage medium. The method comprises the following steps: obtaining sample voice information and sample text information; processing the sample voice information through an acoustic model in an initial recognition model to obtain acoustic feature information; processing the acoustic feature information and the sample text information through a translation model in the initial recognition model to obtain a first predicted translation result and a second predicted translation result respectively; training the initial recognition model based on the first predicted translation result and the second predicted translation result to obtain a target recognition model; and the K vector and the V vector in the acoustic model are spliced with a prefix vector and / or the K vector and the V vector in the translation model are spliced with a prefix vector. The training efficiency of the target recognition model is improved.

The invention discloses a bus passenger getting-off station prediction method and system based on rank learning, and the method comprises the steps: obtaining bus card swiping data and weather data, carrying out the data preprocessing and data normalization processing, and constructing a normalized to-be-predicted line feature matrix; constructing a sorting learning model according to the normalized feature matrix of the to-be-predicted line; and through a LambdaMART algorithm, carrying out feature prediction, index screening and parameter adjustment processing on documents in the sorting learning model to obtain a bus passenger getting-off station prediction result. According to the method, the spatial relationship between the stations can be utilized more effectively, so that the prediction precision and stability are improved. The bus passenger getting-off station prediction method and system based on sorting learning can be widely applied to the technical field of machine learning.