41results about How to "Improve model performance" patented technology

The invention discloses a high-speed rail system sliding layer peak displacement envelope curve and rail direction irregularity prediction method considering seismic oscillation and pier height uncertainty and application. The method comprises the following steps: constructing a multi-mode hybrid neural network model TransTCN embedded with field priori knowledge; feature extraction is carried out on multi-modal input features of a time sequence based on seismic oscillation fusion structure features and a space sequence for outputting peak displacement based on an elastic response spectrum method; the pier height features and the seismic oscillation features are subjected to dimension fusion to serve as time sequence features, and a time dependency relationship is processed; processing a spatial dependency relationship by taking an elastic response spectrum analysis result as a spatial feature; and jointly inputting the extracted features into a constructed multi-mode hybrid neural network model TransTCN, and realizing prediction of the peak displacement envelope curve and rail direction irregularity of the sliding layer of the high-speed rail system under the consideration of seismic oscillation and pier height uncertainty. The prediction precision is improved, the training sample demand is reduced, the regional high-speed rail system damage is rapidly evaluated, and the practical value is good.

The invention discloses an on-load tap-changer fault diagnosis method based on lightweight YOLO11, and is used for solving the problem of low mechanical fault diagnosis efficiency of a traditional on-load tap-changer. Comprising the following steps of: converting vibration signals in a motor starting stage and a change-over switch action stage into high-resolution time-frequency images through continuous wavelet transform (CWT) so as to comprehensively reflect time and frequency characteristics of the signals; starNet is adopted to replace a YOLO model backbone network, and the feature extraction capability is enhanced while the network compactness is kept through star operation; and designing an efficient joint convolution detection head EJCD, and realizing model lightweight and multi-scale target detection through grouping normalization and convolution parameter sharing. According to the invention, through combination of the time-frequency image representation capability of continuous wavelet transform and an efficient target detection framework of deep learning, the limitation of a traditional method in non-stationary signal processing is overcome, and the diagnosis efficiency and the model performance are significantly improved.

The invention discloses a strawberry internal quality and maturity nondestructive testing method based on an MPCD-Mama model. The strawberry internal quality and maturity nondestructive testing method comprises the following steps: acquiring a strawberry sample, acquiring a hyperspectral image by using a hyperspectral reflection imaging system, measuring SSC, PH and hardness, and marking maturity; correcting the hyperspectral image, performing spectrum extraction to form a data set, and dividing the data set into a training set, a test set and a verification set; the method comprises the following steps: constructing an MPCD-Mama model, training the MPCD-Mama model by using a training set, testing the performance of the MPCD-Mama model by using a verification set on the basis of a cross validation method, predicting related indexes on the basis of the pre-trained MPCD-Mama model, generating heat maps of corresponding components at different positions of strawberries, and displaying the distribution conditions of the different components. According to the method, nondestructive testing of multiple quality parameters of strawberries is achieved, compared with a traditional method, the prediction precision is remarkably improved, and technical support is provided for nondestructive testing of the quality of the strawberries.

The invention relates to an ECG anomaly detection method based on AE-GAN, and belongs to the field of artificial intelligence. According to the method, AE-GAN is constructed and trained for ECG anomaly detection, and a training network model is divided into an encoder training stage, the AE-GAN is only trained on a normal ECG sample, and an encoder, a generator and a discriminator are optimized by using confrontation loss and reconstruction loss; the encoder includes a channel attention module and a memory module. In a classifier training and detection stage, the AE-GAN extracts low-dimensional features from a real sample by using a trained encoder, adds weak random Gaussian noise to generate a pseudo normal sample, and adds strong random Gaussian noise to generate a pseudo abnormal sample; the classifier is trained by using a binary cross entropy loss function so as to distinguish normal samples from pseudo abnormal samples; during testing, the trained classifier uses features extracted by the encoder to detect whether data is abnormal or not.

The invention discloses an automatic data analysis system and method based on artificial intelligence, and relates to the technical field of data management, and the method comprises the following steps: S1, data collection and preparation; s2, data annotation and classification; s3, feature selection and extraction; s4, data standardization and normalization are carried out; s5, model selection; s6, training and optimizing the model; s7, performing model evaluation; s8, model interpretation and visualization are carried out; s9, generating an automatic report; and the system is constructed based on the steps. According to the automatic data analysis system and method based on artificial intelligence, by setting a classification and labeling model, a training optimization mechanism and a dynamic model evaluation method, the efficiency and precision of automatic data analysis are remarkably improved, and the quality of data preprocessing is improved; meanwhile, through self-adaptive hyper-parameter adjustment and optimization, the efficiency of the training process and the model performance are optimized; the dynamic evaluation mechanism ensures that the model is more accurate in performance under different data characteristics.

The invention discloses a livestock health state monitoring method, apparatus and device, and a storage medium. The livestock health state monitoring method comprises the steps of obtaining sensor monitoring data in a livestock factory; constructing a sample data set according to the sensor monitoring data; and performing model pre-training and model tuning training on a pre-constructed monitoring network model by adopting the sensor sample data until a preset model training ending condition is met, and taking the trained monitoring network model as a livestock health monitoring model. By means of the technical scheme, the model performance of the livestock health monitoring model applied to different livestock factory areas is improved, and meanwhile the training efficiency of model training is improved.

The invention provides a federated learning-based chip manufacturing process parameter optimization method and system, and relates to the technical field of chips, and the method comprises the steps: collecting multi-factory process parameters and yield data, screening key parameters through wavelet transform and dual evaluation, and constructing a deep Q network model; ethnic group division is carried out based on process trajectory similarity, differential privacy and homomorphic encryption are adopted to protect data security, and ethnic group optimization is realized through an attention mechanism and knowledge distillation; and setting a third-level time window monitoring yield, and dynamically adjusting an optimization strategy. According to the method, the chip manufacturing yield can be effectively improved, data privacy is protected, and intelligent optimization of process parameters is realized.

The invention belongs to the field of traffic control systems, and relates to a vehicle driving track prediction system and method based on an improved graph neural network model, and the system comprises an FPGCN model and a Social LSTM model. The FPGCN model comprises a graph convolution layer, a Dropout layer, a differential pooling layer, a feature fusion module and a full connection layer; obtaining node feature vectors from graph structure data through a graph convolution layer, a Dropout layer and a differentiable pooling layer, then dividing and conquering node features, generating global features through a dense attention mechanism and principal component analysis, and generating local features through virtual filling and matrix calculation; after global features and local features are obtained, a feature fusion module is used for fusion, then the features are input into a full connection layer, and the influence score of a historical trajectory diagram on a current state diagram is calculated; integrating the preprocessed driving data and the data of which the score is calculated by the FPGCN model, and inputting the integrated data into a Social LSTM model; according to the system, while the model expression ability is maintained, the parameter quantity is remarkably reduced, the calculation logic is optimized, and the prediction accuracy is improved.

The invention provides a target identification method and system based on multi-source information fusion, and relates to the technical field of low-altitude target detection. The method comprises the following steps: acquiring target radar track data, interception equipment track data and position area information data; based on target radar track data, extracting a first feature in an RCS form dimension, and extracting a second feature in a motion dimension; based on the target radar track data and the track data of the monitoring equipment, determining a frequency spectrum monitoring correlation factor and regional position information features; performing feature fusion on the first feature, the second feature, the spectrum interception correlation factor and the regional position information feature to obtain a target feature, and identifying a target type; and identifying a target threat level based on the target type and the target radar track data. The method and the device are used in a target identification process based on multi-source information fusion, and the technical problem that the target threat degree cannot be accurately identified in a complex environment in the prior art is solved.

The invention discloses a neural network training method and device based on depth time alignment gradient enhancement. The training method comprises the following steps: acquiring input sample data; inputting the input sample data into a neural network model based on depth time alignment gradient enhancement to obtain prediction category probability data corresponding to the category of the input sample data; according to the prediction category probability data and the sample label data, adjusting each parameter of the neural network model based on depth time alignment gradient enhancement to obtain a trained neural network model, the neural network model based on depth time alignment gradient enhancement comprising an input layer, K stages and an output layer, the first stage to the (K-1) th stage each comprise a convolutional network based on depth time alignment gradient enhancement and an auxiliary classifier network, the Kth stage comprises a convolutional network based on depth time alignment gradient enhancement, and K is a positive integer greater than 1.

The invention provides a model distributed parallel training method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: determining a hyper-parameter combination for training a target model in a distributed computing system; generating a particle swarm based on the hyper-parameter combination; the speed of each particle in the particle swarm is determined based on the value distribution probability of each hyper-parameter in the hyper-parameter combination; training the target model based on a distributed training execution script generated by each particle, updating the speed and position of each particle by taking the training throughput performance of the distributed computing system as the fitness value of each particle, and determining the optimal solution of the hyper-parameter combination; and on the basis of a distributed training execution script generated by the optimal solution of the hyper-parameter combination, continuing to train the target model. According to the method and the device provided by the invention, the training efficiency and the model performance of the model are improved in a distributed computing environment.

The invention provides a malicious user identification method and system giving consideration to privacy protection in a social network, and relates to the technical field of network privacy security, and the method comprises the steps: carrying out the structure perception sub-graph segmentation of a social network graph through an METIS algorithm, dividing an original graph into a plurality of sub-graphs, minimizing the number of edges crossing the sub-graphs, and keeping the scale balance of the sub-graphs; constructing a privacy perception GNN of an integrated gating residual attention module, wherein the privacy perception GNN comprises a privacy perception linear layer and a gating residual mechanism; based on differential privacy stochastic gradient descent framework training, combining an adaptive noise scheduling strategy, dynamically adjusting the noise scale according to privacy consumption deviation, and performing closed-loop control budget to obtain a trained model; and malicious users are identified through the trained model. According to the method, the problem of performance reduction caused by fixed noise injection and noise amplification is solved, and efficient and robust identification of malicious users is realized while differential privacy constraints are met.

The present invention relates to a child sleep staging method based on collecting electroencephalogram signals by headband electrodes, belongs to the technical field of sleep staging, solves the problem of poor comfort and coordination when traditional polysomnography monitoring equipment is used on children, and includes the steps of: constructing a Web client and a server; the headband electrode equipment collects electroencephalogram signals of multiple child subjects; the server adopts a new wavelet threshold denoising strategy based on EMD to denoise the electroencephalogram signals to obtain sleep electroencephalogram data, performs data segmentation and elimination on the data, constructs a time-frequency feature map, trains and tests a multi-view hybrid neural network model, obtains a child sleep staging model, uses the model to perform child sleep staging, and displays the child sleep staging results by the Web client. The present invention can improve the comfort and coordination of children in the data collection process, improve signal quality, and effectively improve the classification accuracy and model performance of the child sleep staging model.

The invention discloses a MobileNet-BiLSTM-Attention tool remaining life prediction method based on an improved sparrow algorithm, and aims to solve the problems of accuracy and uncertainty of tool remaining life prediction in an actual machining environment. The method comprises the following steps that a milling experiment platform is built, and multi-channel signals of a cutter under different feeding times are obtained; intercepting a vibration signal stable interval and carrying out noise reduction processing by adopting a moving average filtering method; constructing a MobileNet-BiLSTM-Attention model, inputting the preprocessed signal into the model, and determining key parameters in the model by adopting a sparrow search algorithm (SCSSA) fused with sine and cosine and Cauchy variation so as to give full play to the advantages of the model to realize accurate and efficient prediction of the residual life of the cutter; a confidence interval of a prediction value is calculated by using an adaptive bandwidth kernel density estimation (ABKDE) method, the uncertainty degree of a prediction result is reflected, and a more stable and reliable prediction result is provided.

The invention provides a drug-resistant cell strain of a KRAS inhibitor as well as preparation and application of the drug-resistant cell strain. The preservation number of the drug-resistant cell strain is CCTCC (China Center For Type Culture Collection) NO: C2023226. A drug-resistant cell strain GP2D-R resistant to MRTX1133 is obtained by adopting a method of combining drug concentration increasing impact with drug-resistant monoclonal amplification for induction, a KRAS (G12D) inhibitor drug-resistant animal model is established, the occurrence and development microenvironment of resistance of colon cancer to a KRAS (G12D) targeted drug can be truly simulated by utilizing the animal model, and the drug-resistant cell strain GP2D-R is of great significance to clinical research of colon cancer.

The invention discloses a hybrid model-based power anomaly recognition method, device and equipment and a medium, and the method comprises the steps: obtaining a to-be-recognized power anomaly image, and inputting the power anomaly image into a pre-trained hybrid power anomaly recognition model; respectively performing first image feature extraction and second image feature extraction on the power anomaly image through the first anomaly recognition model and the second anomaly recognition model to generate a first image feature and a second image feature; carrying out feature alignment on the first image feature and the second image feature; and performing feature fusion on the first image feature and the second image feature after feature alignment to generate a fused abnormal feature based on the abnormal level information of the abnormality of the power equipment, identifying the fused abnormal feature, and outputting the abnormal type of the power abnormal image. According to the technical scheme, the recognition precision and robustness of the model are improved.

The application provides a machine translation model generation method and device, comprising: obtaining training data; training a machine translation model based on a pre-stored deep neural network using the training data; the machine translation model is used to translate a text to be translated into a text translation; the machine translation model comprises a memory enhancement adapter layer, a first memory and a second memory; the memory enhancement adapter layer is used to retrieve information in the first memory and / or the second memory and utilize the information; the first memory is used to save source language data obtained by performing reverse translation and forward calculation processing on the training data; and the second memory is used to save target language data obtained by performing reverse translation and forward calculation processing on the training data. The application saves the training data in a form acceptable to the model, reads information from the training data to assist the model when performing a translation task, and achieves the effect of adapting to various translation scenarios while taking into account lower algorithm time complexity and better model performance.

The application provides an end-edge-cloud collaborative hierarchical federated learning training method, device and storage medium, and belongs to the technical field of distributed machine learning and edge computing. The method is applied to a three-layer system including a cloud server, an edge server and a terminal device, and comprises the following steps: the cloud server distributes a global model to each edge server; the edge server dynamically clusters according to the computing and communication capabilities of the subordinate terminal devices, determines differentiated model calculation compression rates and communication compression rates for different clusters, and distributes lightweight models; the terminal device performs local training and uploads parameters; the edge server and the cloud server perform teacher-free online knowledge distillation based on a shared data set to collaboratively update the model; and the cloud server triggers global aggregation according to a dynamic time threshold scheduling strategy to generate a new round of global model. The application effectively alleviates the training blocking problem in a heterogeneous device environment, improves resource utilization efficiency and model performance, and enhances system convergence stability.

The application discloses a double-embedding model hybrid training method, system, device and storage medium, which is applied to the technical field of recommendation systems and comprises the following steps: determining the routing parameters of each classification feature based on the historical access information of the classification features in a training data set; selecting a corresponding embedding table for the classification features input currently according to the routing parameters; wherein the embedding table comprises a first embedding table and a second embedding table; extracting the embedding vector corresponding to the classification features from the selected embedding table, and performing forward calculation and loss calculation of the model based on the embedding vector; and updating the parameters of the first embedding table, the second embedding table and a downstream model according to the result of the loss calculation. The application simulates the embedding hybrid use mode in reasoning in the training stage, enhances the collaborative ability of the two embedding tables, and improves the performance and robustness of the model in a real reasoning scene.

The invention provides equipment for synthesizing data, and relates to the technical field of data synthesis, and the equipment comprises a picture screening module which is used for dividing to-be-labeled data into pictures difficult to label and pictures easy to label through the module, enabling the pictures difficult to label to enter a manual adjustment module, then enabling the pictures difficult to label to enter a model updating module, expanding a training set, and retraining a network, the pictures easy to label enter an automatic labeling module; according to the equipment for synthesizing the data, automatic annotation of pictures can be realized, universality can be realized to a certain extent by updating a model function, and the efficiency of image annotation is greatly improved; the pictures can be screened more comprehensively through a proposed query strategy, so that the required model performance can be achieved with a smaller number of annotation samples; by providing an automatic labeling method based on instance segmentation, the problem that overlapped pictures are difficult to label is solved.

The application relates to the technical field of aerospace and artificial intelligence, and provides a space target orbit intelligent prediction method fusing an orbit dynamics constraint, which comprises the following steps: preprocessing original orbit data of a space target to generate a complete time series data set with continuity and uniform time intervals; constructing a physical information neural network model which represents input and output in the form of time series data, and performing differential calculation on an output sequence by using an automatic differentiation mechanism; designing and introducing a loss function with an adaptive weighting mechanism to dynamically balance loss terms of network modules and physical modules in a training process; training the physical information neural network model; setting a data subset sampling rate, constructing a multi-scale data scene, and evaluating performance of the physical information neural network model under different data scale conditions. The application can improve modeling capability for a long-time orbit evolution process and realize dynamic adjustment of weights of various loss terms.

This invention relates to the field of remote sensing image processing technology, and in particular to a method for extracting urban green spaces from collaborative multi-source remote sensing imagery. The method includes acquiring multi-source remote sensing image data to construct an urban green space information map, preprocessing the images and extracting features to generate image feature vectors, fusing them through semantic mapping to generate a comprehensive feature vector, training a green space extraction model for small-sample scenarios based on a transfer learning framework, and using real-time imagery to determine green space types and update the map. This invention can improve the accuracy of urban green space extraction, adapt to small-sample scenarios, and achieve dynamic monitoring and information improvement.

The present specification discloses a model training method and a point cloud segmentation method based on an attention mechanism. The method comprises: obtaining a training sample, wherein the training sample includes support set point cloud data and query set point cloud data; inputting the training sample into a semantic segmentation model, determining feature data corresponding to the training sample through a feature extraction network; transposing the feature data to obtain transposed feature data, weighting the transposed feature data through a first attention network to obtain a first weighted feature, and weighting the feature data through a second attention network to obtain a second weighted feature; determining a target feature based on the first weighted feature and the second weighted feature; semantically segmenting the query set point cloud data based on the target feature corresponding to the support set point cloud data and the target feature corresponding to the query set point cloud data, determining a loss value based on the deviation between the semantic segmentation result and its label, and performing model training.

The invention provides a multi-agent collaborative machine learning process automatic generation method and system, and the method comprises the steps: A1, obtaining a target data set from a code warehouse, obtaining the feature information of the data set, and generating a data analysis result; a2, according to the data analysis result, selecting a corresponding process template as prior information, and then scheduling generation of a plurality of agents and obtaining an agent generation result; step A3, verifying and integrating the agent generation results, and forming and executing a complete machine learning process; and A4, after the machine learning process is successfully executed, searching optimal hyper-parameter configuration and outputting a final model result. According to the method, automatic generation of the end-to-end machine learning process under multi-agent cooperation can be realized, and the success rate of process generation, the model performance and the robustness of the system are remarkably improved.

The invention belongs to the technical field of burn personalized skin scaffold systems, and particularly relates to a burn personalized skin scaffold design method and system fusing 3D printing and an adaptation algorithm, and the method and system carry out multi-modal data synchronous collection, modal noise reduction and standardized safety storage. A high-precision clock module is additionally installed for OCT equipment, a temperature sensor, a pH biosensor and a fluorescence imaging module, unified acquisition software is established to set a corresponding sampling frequency, multi-equipment synchronous acquisition is realized through single triggering, and a fast expansion random tree and a particle swarm optimization algorithm accelerated by a GPU are adopted; an edge calculation module is integrated at a printing equipment end, a real-time data processing framework is adopted, a physical and digital twin system is established to optimize printing parameters to perform closed-loop control, and testing is performed on a bionic skin model containing a dynamic fluid circulation system and a mechanical stretching device.

The invention discloses a code reconstruction method and device, equipment, a medium and a chip, and belongs to the field of artificial intelligence. Comprising the steps of generating a UML class diagram according to a call chain of a source code of a to-be-upgraded application; the UML class diagram is a code class relation diagram meeting the reconstruction requirement of the source code; performing function marking on a plurality of functions contained in the source code to obtain function marking information; generating at least one code reconstruction request according to the source code, the reconstruction demand of the source code and the function mark information; a target code is obtained through a pre-trained code reconstruction model according to the code reconstruction request, and the code reconstruction model is obtained through training based on a GPT large language model: a target code framework is obtained through the pre-trained code reconstruction model according to the UML class diagram and the function marking information; and splicing the target code with the target code framework to obtain a reconstructed code of the source code.

The present invention provides a method for knowledge graph completion, aiming to improve the accuracy and effectiveness of knowledge graph completion. The method first obtains and expands the triple data of the knowledge graph, and supplements the data set by generating inverse triples. A variety of negative sample generation techniques are combined to improve the model's ability to identify incorrect links, and a dual text encoder is used to perform text embedding learning on entities and relationships in triples. A deformable adaptive attention module is proposed to capture potential features of text features. The triple similarity score is obtained by calculating the cosine similarity score of the text embedding and adding the path reasoning gain score. The information noise contrast estimation loss function with margin is used to optimize the text embedding, which effectively improves the completion effect. Finally, the maximum similarity matching of triples is completed using the trained knowledge graph completion model. The method of the present invention can capture complex semantic associations and enhance the accuracy and robustness of knowledge graph completion.

The invention discloses a defect detection model construction method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a plurality of training samples; wherein the training sample comprises a copper bar chain sample image and a real defect identifier corresponding to the copper bar chain sample image; processing the copper bar chain sample image according to the feature extraction sub-model and the feature dimension reduction sub-model to obtain a to-be-processed feature, and updating the training sample based on the to-be-processed feature; processing the training sample based on the deep learning sub-model to obtain a third loss value; correcting the model parameter based on the third loss value to obtain a classification sub-model; and constructing a defect detection model based on the feature extraction sub-model, the feature dimension reduction sub-model and the classification sub-model, and carrying out copper bar chain surface defect detection on the to-be-processed copper bar chain image based on the defect detection model. According to the technical scheme, the defect detection model is obtained based on the training sample and multiple algorithms, so that the model performance and generalization ability of the defect detection model are improved.

The invention belongs to the technical field of human body posture estimation, and discloses a lightweight posture estimation method based on a frequency domain attention mechanism, and the method comprises the steps: compressing input feature information through grouping convolution, and then restoring the compressed feature information through three attention modules. Firstly, the frequency attention module uses the filtering module to extract different frequency components, important frequency components are selected through softmax normalization key values, and then the channel and space attention module better captures global information and accurately positions key features. The calculation complexity of the frequency domain attention mechanism module provided by the invention is linearly related to the image resolution, the calculation complexity is smaller, the problem that the calculation amount is rapidly expanded along with the increase of the resolution can be solved, and meanwhile, the module can reduce information loss caused by packet convolution feature compression, so that the accuracy of the algorithm is improved. Therefore, the human body posture estimation accuracy in the lightweight scene is further improved.

The embodiment of the invention provides a distributed training aggregation method for a vehicle-mounted terminal model based on cloud side-end cooperation, and belongs to the technical field of computing, and the method specifically comprises the steps: building a cloud side-end three-layer Internet of Vehicles network system architecture; the vehicle-mounted terminal collects a local data set; the upper layer issues the initial model; the intermediate layer constructs a target problem, converts the target problem into an integer linear programming problem and solves the integer linear programming problem; locally training the initial model and sending the trained model to an intermediate layer; asynchronous weight aggregation is carried out by adopting a hysteresis interference sensing method; returning the aggregated model, and carrying out global model asynchronous aggregation by the cloud central server by adopting a lag interference sensing method to obtain an updated global model; and after aggregation of all global rounds is completed, the upper layer issues a final global model to the middle layer, and then the middle layer forwards the final global model to all the vehicle-mounted terminals. Through the scheme disclosed by the invention, the model distributed training efficiency on the terminal equipment is improved, and the communication overhead in the training process is reduced.