81results about How to "Increase training speed" patented technology

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.

The application provides an optimization deployment system and method based on a MAD-GAN algorithm, and belongs to the technical field of artificial intelligence.The system comprises a bus control unit, a discriminator and a generator; demand generation samples are divided into k modes, and the system is set to a default state; the generator monitors the bus state in real time, generates a random noise vector to obtain sample data; the sample data is written into a half-duplex generator data bus; the discriminator monitors the bus state in real time, reads the data in the half-duplex generator data bus, and reads the comparator parameters and the softmax parameters to obtain an evaluation value and a perception value of the generated sample; after gradient operation on the evaluation value, the evaluation value is cached, and the perception value of the generated sample is written into a half-duplex gradient data bus; after updating the discriminator parameters, the next cycle period is entered.The application can significantly accelerate the training speed of the MAD-GAN algorithm, has a higher requirement for storage capacity, and improves the model friendliness.

This invention discloses a target recognition method, docking system, and docking method in a space environment. By adjusting the output intensity of the sunlight simulator's light source and the position and angle of the simulated work equipment, different types of simulated target images are obtained, forming a target photo library. The simulated target images are processed into small target images and then placed into a small target image library. This small target image library is merged with the target photo library augmented by GAN data, and divided into training and testing libraries proportionally. An SA-YOLOX network is constructed. The SA-YOLOX network is trained using the training library to obtain an SA-YOLOX target detection model. The augmented target photo library is tested using the SA-YOLOX target detection model and the testing library to obtain test results. Based on the test results and model metrics, an evaluation result is output. This invention effectively overcomes the performance index problem of the YOLO series algorithms in solving the special situation of high-contrast image target recognition in complex space environments.

The application provides a traffic classification method and system based on federal semi-supervised learning. The method comprises the following steps: constructing an unlabeled traffic dataset and a labeled traffic dataset; a center server decomposes a global model into supervised learning parameters and unsupervised learning parameters and initializes the parameters; the parameters and an auxiliary agent are sent to each client; each client performs unsupervised training based on the supervised learning parameters, the unsupervised learning parameters and the auxiliary agent by using a local unlabeled traffic dataset, uploads the difference between the unsupervised learning parameters to the center server; the center server aggregates and updates each unsupervised learning parameter; supervised training is performed by using a local labeled traffic dataset, and the difference between the supervised learning parameters and the difference between the unsupervised learning parameters are sent to each client; a new auxiliary agent is obtained based on nearest neighbor search, and the new auxiliary agent is sent to each client when a set sending condition is met; the two steps are iteratively executed until a stop condition is met.

The application discloses a non-intrusive load decomposition method based on wavelet decomposition and an improved neural network, mainly uses a non-intrusive load identification device to collect power consumption information, carries out pretreatment operation on the collected data information, extracts load characteristics through four-layer wavelet decomposition, and then inputs the load characteristics into an improved neural network for training and learning; the improvement mainly lies in that the network is multi-scale and multi-input, the training process is multi-resolution, and the network is divided into four parts according to the wavelet decomposition layers and is trained respectively, is spliced again finally, then a large amount of data is used to adjust network parameters, a non-intrusive load decomposition model is perfected, a load decomposition task is completed, and the non-intrusive load decomposition result is analyzed. Through the application, the time-frequency domain can be combined, the problem that important information can be lost, data characteristic utilization is low and a multi-state process electric appliance is difficult to identify when load data is too much can be solved.

The application discloses a patch inductance surface defect detection method and system based on token fusion, and the method comprises the following steps: acquiring a patch inductance defect data set; introducing a token fusion module based on a VisionTransformer network model to perform optimization processing, so as to obtain an optimized VisionTransformer network model; performing surface defect detection processing on the patch inductance defect data set based on the optimized VisionTransformer network model, so as to obtain a patch inductance surface defect detection result. The system comprises an acquisition module, an optimization module and an evaluation module. Through the use of the application, the token fusion module is introduced to realize a patch inductance surface defect rapid and accurate detection module. The application can be widely applied to the field of image recognition technology.

The application relates to a liver cancer CT image segmentation method based on channel attention, a segmentation network for liver and tumor regions is constructed, residual modules are used to replace ordinary convolution layers on the basis of a U-Net encoding-decoding structure, the training speed and precision of the network are improved, an attention mechanism is added in the decoder stage, the importance and mutual dependence of different channels in a feature map are fully considered, feature maps output by each stage of the decoder are up-sampled to the size of an original image through a deep supervision module and are spliced with the feature map output in the last stage, feature loss is prevented, two same segmentation networks are used for segmenting a liver region of interest and segmenting tumor cells respectively. The method has good segmentation effect without significantly increasing the parameter amount; from the clinical point of view, liver and tumor regions in an abdominal CT image are accurately segmented, a good auxiliary role is provided for doctor diagnosis, and the diagnosis efficiency is improved.

This invention discloses a method, system, device, and storage medium for predicting oil pipe corrosion rate based on a PCA-PSO-SVR hybrid model, specifically including the following steps: S1 Collecting corrosion detection data and operating condition parameters of the oil pipe to form a dataset; S2 Preprocessing the dataset; S3 Using the PCA model to perform dimensionality reduction on the dataset and extracting the main features affecting the corrosion rate; S4 Initializing the parameters of the PSO model; S5 Optimizing the parameters of the SVR model using the PSO model; S6 Constructing a corrosion rate prediction model based on the optimized SVR model; S7 Inputting the preprocessed dataset into the corrosion rate prediction model to obtain the prediction result and complete the prediction.

The embodiment of the application relates to the technical field of image processing, and discloses an image comparison method, system, electronic equipment and computer readable storage medium, the method comprising: respectively pre-processing and adaptively binarizing a test image and a template image corresponding to the test image to obtain a first image and a second image; performing target detection on the first image, respectively removing a detected target region in the first image and a region in the second image with the same position as the target region to obtain a third image and a fourth image; extracting feature points in the third image and the fourth image and performing feature point matching to determine matched feature points; performing global binarization on the pre-processed test image and the template image and performing difference processing to obtain a difference image; and obtaining an image comparison result based on the matched feature points and the difference image, so that different elements between images are efficiently identified, elements that do not need to be focused on are excluded, and the workload of a tester during rechecking is reduced.

The invention provides a method for optimizing an artificial intelligence model based on a tensor structure, and the method comprises the steps: replacing a linear mapping weight matrix in an initial artificial intelligence model with the tensor structure, and obtaining a target artificial intelligence model, the tensor structure comprising a target tensor and / or a target tensor network, the target artificial intelligence model is used for processing a target task, and the target task comprises at least one of natural language processing, logic and mathematical reasoning, code programming and multi-modal content generation. Structured compression of the artificial intelligence model is achieved through the target tensor and / or the target tensor network technology, the calculation complexity is effectively reduced while the model parameter quantity, hard disk occupation and video memory occupation are greatly reduced, and the reasoning and training speed of the model can be remarkably increased.

The present application relates to a kind of ground airglow gravity wave image fast determination method and system, method includes: collecting real-time ground airglow image;After the real-time ground airglow image is preprocessed, input layered transfer learning model, output gravity wave existence probability, and save wave image, wherein, the layered transfer learning model is obtained by segment training to ground airglow image dataset, the layered transfer learning model includes pre-training EfficientNetB3 model, deformable convolution layer and classification head.The present application solves the problem that ground airglow imager is difficult to automatically distinguish gravity wave in short-term deployment in real time, reduces artificial expert discrimination workload, while avoiding the problem of overfitting low accuracy of traditional machine learning method in short time small amount of data training.Reach the purpose of rapid deployment, efficient identification.

This invention relates to the field of battery state prediction technology and discloses a method for predicting the health status of UAV power batteries. The key technical points of this method are: data acquisition and quantum encoding; quantum feature extraction; construction of a quantum-heuristic deep learning model; model training and optimization; and health status prediction and assessment. Through quantum encoding, a quantum-heuristic model architecture, and a dynamic optimization mechanism, this method overcomes the limitations of traditional methods in modeling complex nonlinear relationships and local optimization, providing a high-precision and robust solution for the health management of UAV power batteries, and has significant engineering application value.

This application discloses a federated large language model training method, apparatus, and device. The method includes: constructing an initial mask matrix for each original model parameter in the initial large language model; splitting the initial mask matrix to obtain initial mask blocks based on the current pruning round; determining the current mask block corresponding to the initial mask block based on the mask reconstruction dataset, a predetermined sparsity rate, the original model parameter matrix corresponding to the initial mask block, and the inverse Hessian matrix of the original model parameter matrix, and determining the current model parameters corresponding to each non-zero, non-target element in the current mask block to obtain the current model parameter matrix; determining the pruned large language model based on the current model parameter matrix corresponding to each pruning round; and optimizing and training the pruned large language model based on a first optimized dataset to obtain the target large language model. This application can perform reasonable and accurate model pruning, simplify the model structure, and improve the model training speed.

This invention discloses a method, apparatus, electronic device, and storage medium for training an intelligent agent policy network. The method includes: determining the linear score of a first training sample; inputting the current first training sample into a state-action value network, a state value network, and a policy network respectively, and determining a first actual output, a second actual output, and a third actual output; determining a first loss value based on the second actual output, the cumulative reward at a first historical time, the linear score, and a first loss function, and adjusting the parameters of the state-action value network; determining a second loss value corresponding to the state value network, and adjusting the parameters of the state value network accordingly; determining a third loss value corresponding to the policy network, and adjusting the parameters of the policy network based on the third loss value, to obtain the target policy network. This technical solution improves the training effect and speed of the policy network, enabling more accurate and convenient training of the target policy network.

The application relates to the computer technical field and discloses a modeling method of a statistical mixed model in a big data distributed scene. The method comprises the following steps: distributing and storing data shards; initializing model parameters; iteratively performing an expectation step and a maximization step; the expectation step is scheduled to a GPU node to perform parallel calculation on posterior probability; the maximization step is scheduled to a CPU node to aggregate statistics and update parameters, perform component merging / deletion, and perform convergence judgment; meanwhile, a memory reuse mechanism based on reference counting and scope analysis is adopted to reduce redundant data transmission. Through heterogeneous task scheduling and memory collaborative optimization, the training speed, resource utilization rate, and model self-adaptation capability are improved.

The application discloses a kind of multi-unmanned plane local dynamic obstacle avoidance method, device and storage medium, wherein method includes: based on Markov decision process, formalization modeling is carried out to distributed multi-unmanned plane dynamic obstacle avoidance problem;Obtain point cloud data by laser radar carried on each unmanned plane, point cloud data is processed, and obtain obstacle information;Deep neural network is constructed, to complete the mapping of observation input to action output and the method of network updating;Wherein observation input is obtained obstacle information;Based on ORCA algorithm, deep neural network is trained, and the model after training is used for multi-unmanned plane local dynamic obstacle avoidance.The application uses dynamic obstacle avoidance algorithm based on geometry method and deep reinforcement learning method as the obstacle avoidance method of multi-unmanned plane system, effectively solves multi-unmanned plane local path planning and obstacle avoidance problem.The application can be widely applied in multi-unmanned plane local dynamic obstacle avoidance and path planning field.

This application discloses a federated learning method, apparatus, device, and storage medium, relating to the field of privacy computing technology. Applied to a federated system composed of multiple participants, the method includes: acquiring the local model parameters of each participant in the federated system; iterating Round training through a local base network and an autoencoder's encoding / decoding layer, and determining the loss value corresponding to the completion of multiple epochs in each Round training process using a preset loss function; sending the local model parameters and loss values ​​to a secure aggregation server, receiving the global parameters and convergence status returned by the secure aggregation server, updating the local model parameters according to the global parameters, and stopping Round training when the convergence status meets preset conditions. The technical solution of this application can improve the learning effect of federated learning in scenarios where the participant data is not independent and identically distributed, and also improve training speed.

The invention provides a surface mine live-action modeling system and method. The method comprises the following steps: automatically acquiring oblique photography images and laser point cloud data of a target mine area through an intelligent data acquisition unit; the data is stably transmitted back to a local data processing center through the data communication and transmission unit; performing automatic preprocessing and quality evaluation on the returned data through a data preprocessing and quality evaluation unit to form a quality control closed loop; a live-action three-dimensional model which is high in precision and supports real-time rendering is generated through a live-action three-dimensional modeling engine unit on the basis of a 3D Gaussian technology and by fusing multi-source data and mine semantic constraints; according to the method, the high-precision live-action three-dimensional model supporting real-time interaction is efficiently and automatically generated, and a powerful data basis is provided for mine digital twinning, safety production and intelligent decision making.

The application provides an antenna array surface fiber grating layout method based on a genetic algorithm and a neural network, and relates to the technical field of deformation sensing of an antenna array surface structure and optimization of fiber grating layout. The method analyzes strain sensitive positions of the array surface structure through the genetic algorithm and the neural network, arranges fiber grating sensors at optimal layout positions of the array surface structure, and realizes high-precision deformation sensing of the array surface structure under complex service loads. The method solves the problems of node layout redundancy, high data processing cost and insufficient displacement prediction accuracy in traditional antenna monitoring, and provides technical support for efficient sensing of antenna structure deformation and stable guarantee of electromagnetic performance.

The application discloses an image classification method and device, a storage medium and a computer device. The method is applied to the computer device and comprises the following steps: in a self-defined image classification mode of an electronic album, in response to a triggering operation of a user on an update control of an image corresponding to a to-be-managed image category, updating the image in the to-be-managed image category, calling a training interface, obtaining a first image classification model corresponding to the self-defined image mode through the training interface, the model being pre-trained, and performing transfer training on all images of the to-be-managed image category according to the first image classification model, so as to obtain a first image classification model after transfer training. The classification of the updated image can be matched by using the first image classification model after transfer training, the classification accuracy is improved, finally, the to-be-classified images of the electronic album are classified by using the trained first image classification model, the image classification efficiency and accuracy are improved, and the image searching efficiency is further improved.

The application relates to a video instance segmentation method based on a dynamic convolution decomposition of a lightweight attention mechanism; the method first inputs a video frame, and a backbone network independently extracts a feature map of each frame in the video; then the feature map extracted by the backbone network enters an HQT encoding and decoding module, the position change of an instance in each frame is accurately positioned by combining an output head, and three prediction branches are used to supervise model training; in the application, a convolution layer of an encoding network adopts dynamic convolution decomposition, a more compact model is obtained, model training is easier, the required parameter quantity is greatly reduced, the training speed is improved, and the training time is shortened; a lightweight HQT encoding and decoding module is provided, which helps to reduce model parameters and improve efficiency; the loss function is improved, model training is more stable, the convergence speed and convergence precision are improved, and the problems of imbalance between foreground and background in samples and imbalance between foreground categories under a long tail condition are relieved.

This invention proposes a dynamic routing decision-making method, specifically a TSSDN dynamic routing decision-making method based on the DDPG deep reinforcement learning algorithm. The aim is to design dynamic routing decisions using deep reinforcement learning algorithms based on network dynamic awareness prediction results, thereby reducing the transmission latency of low-priority flows in time-sensitive networks. This invention uses a deep learning algorithm to predict switch queue lengths in real time, and then makes next-hop routing decisions based on the prediction results. Higher prediction accuracy leads to better dynamic routing performance and improved decision-making efficiency. The implementation steps are: 1) Constructing the TSSDN network node architecture; 2) Constructing the network topology; 3) Constructing a topology feature extraction method based on PCA; 4) Constructing a prediction model based on a deep learning algorithm; 5) Constructing a routing decision-making model based on a deep reinforcement learning algorithm; 6) Iteratively training the deep reinforcement learning-based routing decision-making model based on the prediction results. This invention can be applied to scenarios such as telemedicine.

The application discloses a quad-rotor formation obstacle avoidance control method based on an improved DDPG algorithm, and belongs to the technical field of unmanned aerial vehicle formation control. The method adopts an improved DDPG reinforcement learning algorithm to plan an obstacle avoidance path of the quad-rotor. When the improved DDPG reinforcement learning algorithm is executed, the priority weight of each quadruple experience in the experience replay pool is initialized. After the quadruple experience in the experience replay pool is sampled and trained, the priority weight of each quadruple experience is recalculated based on a TD error, and a Sum_Tree structure is updated. The method effectively alleviates the training instability caused by hyperparameter sensitivity, the misleading of policy updating caused by overestimation of Q values, and the problem that key experiences are not sufficiently learned, accelerates the convergence speed of the quad-rotor formation obstacle avoidance training, and improves the obstacle avoidance effect.

The application discloses an aircraft cluster cooperative non-cooperative target capturing method driven by entropy regularization, applies the entropy regularization thought to reinforcement learning, optimizes the updating method of the parameters of an evaluation network and a policy network, so that a UAV can obtain a more optimal maneuvering strategy, and the UAV performs actions according to the maneuvering strategy, and realizes the interception of an attacking aircraft. The aircraft cluster cooperative non-cooperative target capturing method driven by entropy regularization can not rely on a deterministic strategy in the interception process of multiple UAVs, but adopts a random strategy, avoids the training from falling into a local optimal point, and the actions of the aircraft are randomized as much as possible while the aircraft completes the task, so that the exploratory and robustness can be practically improved.

This invention discloses a method for displacement field recognition of cement-based materials based on a one-dimensional convolutional model. The method includes: S1: extracting the displacement field from a speckle image; using speckle image pairs as feature data, and using the lateral displacement field, longitudinal displacement field, and combined displacement field as label data, and normalizing the label data to obtain a sample dataset; S2: constructing a displacement field recognition model for cement-based materials based on a one-dimensional convolutional model; S3: constructing a model loss function containing the design parameters to be optimized by introducing physical loss, and training the displacement field recognition model based on the one-dimensional convolutional model for cement-based materials based on the sample dataset to obtain the optimal displacement field recognition model; and realizing the displacement field recognition of cement-based materials based on the optimal displacement field recognition model. This invention solves the problem of insufficient accuracy and efficiency in the displacement field recognition of cement-based materials by existing methods.

The invention discloses a federated learning classification system and method for protecting privacy of single-cell RNA sequencing data, and relates to the technical field of data privacy classification. The scRNA-seq data are standardized, and features are enhanced through comparative learning and an auto-encoder; the multi-model dynamic collaborative training module supports dynamic switching of models such as a graph convolutional network and a cross-modal Transform according to data features; the causal inference intelligent evaluation module recommends an optimal model through a three-level evaluation system and a deep Q network; the layered compression communication optimization module adopts a three-stage compression strategy to reduce communication traffic; and the security multi-party computing privacy protection module integrates encryption and differential privacy mechanisms to guarantee data security. The single-cell RNA sequencing data federal learning classification method supports multi-model dynamic adaptation, improves classification precision and system universality, reduces communication overhead, adapts to heterogeneous clients, guarantees data privacy, and achieves efficient and safe single-cell RNA sequencing data federal learning classification.

The application discloses a paper medicine box steel seal character recognition method based on an improved YOLOV5 model, and the method comprises the following steps: collecting the steel seal character image of the medicine box by using an image acquisition device; performing image enhancement and other pretreatments on the image and labeling the recognition area; inputting into the improved YOLOV5 model for training to obtain the trained YOLOV5 model; the improved YOLOV5 model comprises the following steps: adding an efficient position attention mechanism CA (Coordinate attention) in the backbone network of YOLOV5, so that the model network can pay attention to a wide range of position information without bringing too much calculation amount, which is helpful to improve the model performance and better locate and identify the target; using SimSPPF to replace SPPF fast pooling layer to improve the training speed; for the dense small target, an additional smaller prior box is used in the model, and a smaller detection head is additionally added; inputting the data set to be recognized into processing to obtain the final prediction result.

Embodiments of the present disclosure disclose a data processing method based on reinforcement learning, an electronic device and a readable medium. The method is applied to an execution node, a policy node and a training node included in a reinforcement learning system, and a specific implementation of the method comprises: performing environment simulation by the execution node to obtain an observation result; performing policy updating by the training node according to a training sample and a model to be trained, wherein the training sample is transmitted from the execution node to the training node through a sample transmission stream; the policy node performs policy inference according to the observation result and updated policy information, wherein in a non-in-line inference mode, the execution node and the policy node perform data transmission on the observation result and an action to be executed through an inference transmission stream; and the execution node executes the action to be executed to generate an updated observation result. The implementation realizes effective scheduling of large-scale computer hardware resources, meets the training requirements for complex reinforcement learning tasks, and improves the training speed.

The application discloses a polar code construction method based on genetic algorithm acceleration convergence, which comprises the following steps: step 1, setting parameters; step 2, initial genetic algorithm population; step 3, selecting two parents from the population by using a roulette algorithm; step 4, crossing the two parent information bits to generate offspring; step 5, mutating the unlocked information bits of the offspring by using the non-locked frozen bits; step 6, adding the mutated offspring to the population, selecting the optimal one according to the fitness, updating the population and recording the population information; step 7, repeating steps 3 to 6 until the number of genetic iterations reaches a number of spans, judging whether the optimal fitness decreases to a limit or not, if yes, reducing the number of locked bits, temporarily releasing the channel, repeating steps 3 to 6 until the number of genetic iterations reaches a minimum number of spans, if the optimal fitness decreases and the temporarily released channel changes, updating the channel locking condition according to the temporarily released channel, if no, directly repeating steps 3 to 6; and step 8, ending when the genetic iteration stopping condition is met. The application can solve the problem of slow convergence speed by dynamically locking the channel.

The application discloses a matrix decomposition method based on secure aggregation and key exchange, and provides a new idea for enhancing data security of federated learning by performing secure aggregation on the gradient of an item matrix I of matrix decomposition under a federated learning framework; the training sample of a recommendation model (namely, a federated learning model) is efficiently utilized by using the local and securely aggregated gradient, so that the user data is ensured not to leave the local, and meanwhile, the recommendation model training process is made more secure; the gradient is masked and added with noise, so that the leakage of source data information caused by exposure of the real gradient is effectively avoided; and the gradient aggregation mode based on secure aggregation is provided, and compared with the homomorphic encryption technology adopted in the background art, the gradient encryption and decryption have lower calculation complexity and faster calculation speed, and the training speed of the recommendation model is improved.