347results about How to "Accelerated training" patented technology

The invention discloses a robot agent reinforcement learning training method in a complex scene, and the method comprises the steps: constructing a reinforcement learning algorithm environment and a state transition model which comprise a barrier, a dynamic target, environment constraints and other complex scenes and are suitable for a multi-agent system; an independent neural network is constructed for each robot agent by adopting a reinforcement learning algorithm; taking vectors of a given state, time sequence information and historical state information of the intelligent agent as input of the state transition model, and outputting a state transition probability and reward information of a prediction state; the action strategy and the action value output by the neural network are executed in the state transition model, and a new prediction state and reward information are generated; through adjusting hyper-parameters of the neural network, a large-scale distributed training architecture is used to carry out cooperative training on the decision network and the evaluation network, parallel learning and cooperative decision of multiple agents are realized, and performance and generalization ability of the agents in a complex scene are improved.

The invention belongs to the technical field of image classification neural networks, and discloses a lightweight image classification neural network architecture system based on knowledge distillation. The invention provides a lightweight image classification neural network architecture system based on knowledge distillation. The lightweight image classification neural network architecture system comprises a data set preprocessing module, a teacher model preprocessing module, a subnet model construction module, a personalized training module, a real loss calculation module, a distillation loss calculation module, a total loss calculation module and a student model construction module. The invention provides a lightweight neural network architecture system based on knowledge distillation, and designs a brand new training mode by using the superiority of knowledge distillation for teacher model weight extraction: the knowledge distillation of teacher and student models is combined with a deep compression method; the system for constructing the lightweight network can be used for training various complex image classification neural networks so as to improve the network performance.

The invention discloses a multi-agent robot cooperative control method based on distributed reinforcement learning, and the method comprises the steps: building a multi-agent reinforcement learning algorithm environment, and defining a state space, an action space and a reward function; constructing a distributed reinforcement learning training framework which comprises a sampling module, a training module and a scheduling management and storage module; based on an actor-commentator architecture, constructing an independent actor network and a commentator network for each robot agent; through a distributed training framework, sample data sharing of the sampling module and neural network parameter sharing and cooperative training of the training module are realized; and finally, loading the trained neural network into a multi-agent robot system, and carrying out cooperative control according to environment changes and task requirements. Therefore, the robot intelligent body of each node can share experience, and cooperative training is shared through neural network parameters, so that a cooperative control strategy is realized, and the overall performance and robustness of the system are improved.

The invention discloses a data communication method and system based on information security, and belongs to the technical field of data communication, and the data communication method comprises the stages of preparation, transmission, intrusion detection and defense, receiving verification and cleaning ending. The two parties authenticate and exchange keys during preparation; encrypting data by using a symmetric key and an AES algorithm; automatic defense is performed after intrusion detection; the receiver verifies the data; and cleaning the resources at the end, and performing log optimization on the system. According to the invention, through innovatively improving the intrusion detection and defense method, based on real-time network flow monitoring, intrusion analysis based on machine learning and an automatic defense mechanism, the detection accuracy and efficiency are greatly improved, and false alarms and missing alarms are reduced; defense capability and flexibility are enhanced, attacks are intelligently blocked and isolated, and cooperative emergency response is rapid; data communication security technology progress and expansion are promoted, and network security is powerfully guaranteed.

The invention provides a natural language control method for a humanoid robot. By constructing the simple skill library and the complex skill library of the robot, the dependence on a large amount of operation data in a specific field is effectively reduced. According to the method, a large language model is used as a base of a task planner, and the capabilities of scene learning, semantic understanding and common sense reasoning of the task planner are brought into full play. By introducing a visual model in the field of computer vision, high-level semantic information and low-level spatial information of the environment can be obtained, and an accurate scene graph is constructed to serve as environment state representation. According to the voice interaction interface constructed by the invention, the naturalness of man-machine interaction is improved, so that a user can more conveniently communicate with the robot and input instructions, the user experience is effectively improved, and popularization of the humanoid robot in an actual application scene is facilitated.

The invention discloses a high-resolution remote sensing image change detection method based on multi-scale convolution decoding, electronic equipment and a storage medium, and belongs to the technical field of remote sensing image change detection. In order to meet the high-resolution remote sensing image change detection requirement, the method comprises the following steps: constructing a remote sensing image data set; a first stage and a second stage are formed based on a CNN module, a third stage and a fourth stage are formed based on a MambaVision module and a Transform module, and four dual-time-phase multi-scale feature maps are respectively obtained through processing of the four stages; constructing a time feature interaction module TFIM; a multi-scale convolution decoder EMCAD is constructed; and constructing a loss function, carrying out training, testing and evaluation by using the constructed remote sensing image data set, and applying the obtained trained optimal model to high-resolution remote sensing change detection. According to the invention, the detection capability of tiny changes in a high-resolution image is improved.

The intelligent factory data acquisition platform comprises a data sensing layer, a multi-protocol transmission layer, a data fusion layer and an application service layer, the data sensing layer is used for collecting equipment state and environment parameter multi-source heterogeneous data through various sensors, preprocessing the collected data and transmitting the preprocessed data to the multi-protocol transmission layer; the multi-protocol transmission layer is used for receiving the preprocessed data transmitted by the data sensing layer, packaging the data through a multi-protocol data transmission unit, then carrying out multi-protocol transmission, and outputting the transmitted data to the data fusion layer; the data fusion layer is used for receiving the multi-source heterogeneous data transmitted by the multi-protocol transmission layer, performing feature extraction and fusion on the heterogeneous data by adopting a self-correcting convolutional recurrent neural network, and outputting fused feature representation to the application service layer; and the application service layer is used for receiving the fusion features output by the data fusion layer, and performing equipment health diagnosis, process parameter optimization and energy consumption statistical analysis intelligent application services based on the fusion features.

The invention belongs to the technical field of automatic driving, and discloses an autonomous driving decision near-end strategy optimization method based on distributed exploration enhancement. According to the invention, a distributed architecture of the reinforcement learning decision module of the autonomous vehicle is designed, through parallelization and distributed processing, multiple agents can execute strategy training in different environments at the same time, and strategy updating and empirical data are shared in real time. In addition, the distributed architecture allows the method to operate in a larger-scale data set and a complex environment, and the adaptability and generalization ability of autonomous driving decision making are improved. In addition, a random network exploration (RND) module is also combined, the diversity and effectiveness of exploration are further enhanced, and a near-end strategy optimization algorithm used by the reinforcement learning decision module is perfected. By combining the distributed architecture and the RND module, the training process is remarkably accelerated, the exploration efficiency is improved, the decision-making quality of the automatic driving system is comprehensively improved, and the reliability and efficiency of automatic driving decision-making are improved.

The invention relates to a method for converting Lao characters into phonemes based on transfer learning, and aims to realize efficient phoneme conversion by utilizing a model pre-trained on other languages and finely adjusting limited Lao data. According to the method, the problem of data scarcity can be relieved, the training process of the model can be accelerated, and the generalization ability of the model on new tasks is improved. A more robust Lao character transphoneme model can be constructed through transfer learning, and a foundation is laid for subsequent speech synthesis. As Thai with high language similarity with Lao is also a low-resource language and cannot be used as a pre-training language, Chinese is selected as the pre-training language. A Chinese open source data set is firstly used for training a Chinese G2P model based on a Transform architecture, decoding end parameters of the model are initialized to a student model, fine tuning training is performed by using a Lao data set with the scale of ten thousand, and the accuracy of a Lao character-to-phoneme model is improved.

The invention relates to a training method of a target detection model based on knowledge distillation, and belongs to the field of target detection. Comprising the following steps: constructing a training data set and a target detection model, and training the target detection model based on the training data set to output an inference graph; calculating a first loss function to obtain a trained target detection model, and taking the trained target detection model as a teacher model; simplifying parameters of the teacher model to obtain a student model; obtaining a second training data set based on the training data set, training the teacher model and the student model based on the second training data set, and respectively outputting prediction labels; and calculating a second loss function to obtain a trained student model as a trained target detection model. According to the method, the optimal distillation temperature and loss coefficient are obtained by adopting the differential mutation algorithm, and the trained target detection model is obtained by training the student model through knowledge distillation, so that the problem of low detection precision caused by the fact that the existing target detection model is trained by a small amount of samples is solved.

The invention discloses an emotional support dialogue system based on a variational Bayesian inverse reinforcement learning strategy, which is characterized by adopting a strategy emotional support dialogue system constructed by an emotional support strategy selection module, a dialogue common sense analysis and reasoning module and a reply generation module, the emotion support strategy selection module performs strategy selection through a dialogue state encoder and a variational Bayesian inverse reinforcement learning strategy selection module according to current dialogue historical information and used strategies; the dialogue common sense analysis and reasoning module generates common sense information analysis and prediction related to the current dialogue; and the reply generation module generates a reply according to the generated content and the dialogue history. Compared with the prior art, the method has the advantages that the information richness and accuracy of reply generation are enhanced, the strategy selection accuracy and the user experience in a strategy emotion support dialogue scene are effectively improved, and the method has good application prospects and commercial development value.

The invention discloses a PVA film gas permeability intelligent detection method based on machine learning, and relates to the technical field of material testing. The method comprises the following steps: acquiring experimental environment data, first detection data, first diffusion theory characteristic data and second detection data, and constructing and training a neural network model; setting application environment parameters, obtaining third detection data, and calculating second diffusion theory characteristic data according to the application environment parameters and the third detection data; and inputting the application environment parameters, the third detection data and the second diffusion theory characteristic data into a neural network model to obtain a gas permeability prediction value. According to the method, the neural network model is constructed and trained, and the gas permeability is predicted through the neural network model, the application environment parameters, the third detection data and the second diffusion theory characteristic data, so that the dependence on a differential pressure method gas permeability experiment is reduced, the time cost of material testing and experiment is reduced, and the test efficiency is improved. And meanwhile, online monitoring and adjustment of the production process are facilitated, and production benefits are improved.

The invention discloses a blood vessel segmentation method and device based on large model fine tuning, equipment and a medium, and the method comprises the steps: obtaining a blood vessel image, carrying out the preprocessing of the obtained blood vessel image, and obtaining a training set; performing fine tuning on a pre-trained image segmentation model by adopting the training set, so that the image segmentation model learns related features of blood vessel segmentation; wherein the low-rank adaptive fine tuning technology is adopted to perform fine tuning on the image segmentation model: two parallel low-rank matrixes are inserted into a Transform layer in the image encoder, the input is processed by the two low-rank matrixes in sequence, and the processing result and the original output are added to obtain the final output. According to the method, the defects of a large model in a blood vessel segmentation task are overcome through efficient parameter fine tuning and a multi-scale feature fusion module. Accurate blood vessel segmentation can be realized under limited annotation data, the difficulty and cost of medical image segmentation tasks are greatly reduced, and the method can be widely applied to the technical field of medical image processing.

The invention belongs to the field of integrated circuit design, and particularly relates to a chip layout optimization method, system and device based on improved reinforcement learning. The method comprises the following steps: S1, constructing a layout optimization model comprising a feature coding module, a feature fusion module, an experience pool, a strategy network and a value network; the feature coding module is used for extracting a position mask, a line mask and a view mask of the netlist file; the feature fusion module extracts local features and global features and generates comprehensive features; the experience pool, the strategy network and the value network form a reinforcement learning framework. S2, using a large number of random netlists as sample data for training a layout optimization model; and S3, inputting a chart file to be optimized into the trained layout optimization model, outputting an optimized netlist file by the layout optimization model, and generating a corresponding optimized circuit layout chart according to the netlist file. The chip layout optimization method solves the problems that an existing chip layout optimization method is low in efficiency and insufficient in generalization.

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

The invention discloses a non-invasive blood pressure prediction method based on signal dimension extension and a lightweight network, which comprises the following steps: firstly collecting PPG blood pressure data and ABP data, then denoising the collected PPG blood pressure data and ABP data, and cleaning the denoised data. The processed data is used for data signal dimension expansion, a PPG blood pressure measurement model is constructed, and the PPG blood pressure measurement model comprises a basic convolution layer and a residual grouping convolution module Group ResidualBock; and then deploying the constructed PPG blood pressure measurement model into the wearable device to obtain the wearable blood pressure monitoring device. And finally, measuring the processed PPG signal by using wearable blood pressure monitoring equipment to obtain a blood pressure value. According to the method, the PPG signal dimension is expanded, the data processing link is optimized, and the model blood pressure prediction accuracy is improved; and meanwhile, a lightweight algorithm architecture is adopted, so that resource occupation is reduced, the system can be deployed in wearable embedded equipment, and the security of personal privacy data is guaranteed.

The invention relates to a deep learning framework based on multi-modal data fusion and adaptive learning, which comprises a multi-modal data fusion module, an adaptive learning module and a deep learning model module, fusing in different stages of feature extraction by adopting a plurality of fusion strategies to obtain fused features; the deep learning model module is used for performing feature extraction on data of different modalities according to different depth models, performing prediction according to fusion features obtained by the multi-modal data fusion module, and outputting a prediction result; and the adaptive learning module is used for dynamically adjusting hyper-parameters of the depth model by adopting a learning mechanism. Compared with the prior art, the method has the advantages of providing more comprehensive information representation, self-adaptive learning ability, wider application prospect and the like.

The invention relates to a medical image processing technology, and discloses a pan cancer species histopathologic full-slice classification method and device, which is a trained multi-modal pan cancer species classification model based on histopathologic full-slice image data and full-slice diagnosis report pairing data of various tissue parts and various cancer species, and has the advantages of wide application range and high efficiency. And full slices of various pathological tissues can be classified. For classification tasks of different tissue parts, only text prompts (prompts) are used, i.e., a candidate cancer species list is set, and zero-sample full-slice classification can be realized without training. And for rare cancer species, full-slice classification can be carried out. Besides, the model has high expansibility, in order to better adapt to downstream tasks, better classification performance can be achieved by combining slice embedding with a linear classifier for use, freezing a slice embedding part and only finely adjusting the linear classifier, and the requirement for the training data size is low.

According to the model training-oriented gradient data transmission method and the related equipment provided by the invention, the compression ratio is dynamically determined according to the change of the transmission rate by monitoring the transmission rate in real time, so that the gradient data can be intelligently compressed according to the compression ratio, the data transmission quantity is remarkably reduced, and the network burden is relieved; and meanwhile, the transmission rate of each working end is adjusted through congestion control according to the bandwidth of a transmission link between the switch and the parameter server so as to relieve congestion. Under the combined action of the two, the transmission time of the gradient data is remarkably shortened, and the data transmission efficiency is improved, so that the training process of the distributed artificial intelligence model is accelerated, and the overall training efficiency is improved.

The invention discloses a meta reinforcement learning model migration training method and device based on topological transformation, and relates to the technical field of machine learning. Obtaining power data of a plurality of regions, and constructing a region data set; performing data feature comparison on the region data set and a model data set of a preset training model, and determining an updating strategy of the preset training model according to data features; and migrating data in the regional data set into the model data set, and executing an updating operation according to the updating strategy to obtain a target model. According to the method, the regional data set is constructed by collecting the multi-regional power data, the regional data set is subjected to data feature comparison with the model data set of the preset training model to determine the model updating strategy, and the migration training method is effectively applied to model updating. According to the method, the training process is accelerated, the model performance is improved, and the computing resource demand is reduced, so that an efficient and economical model updating mode is realized, and the adaptability and accuracy of the model in power fault positioning are enhanced.

The invention discloses an operation optimization method and system for a power distribution network and a micro-grid group, and a related device, and belongs to the technical field of power grid optimization. The method comprises the following steps: deploying a pre-trained multi-agent optimization model in a power grid environment, acting on the power grid environment, inputting a state space and rewards fed back by the power grid environment into the pre-trained multi-agent optimization model, and outputting an action space of a corresponding agent, taking the action space as a power distribution network and micro-grid group operation optimization scheduling result; according to the method, intelligent agent accelerated training and optimized operation decision-making oriented to the power distribution network and the micro-grid group are realized. The adaptive capability, the transferability and the decision-making capability of multi-agent reinforcement learning are utilized, consumption of renewable energy sources is effectively promoted, and the system operation efficiency is improved while voltage out-of-limit is suppressed.

The invention relates to a landslide disaster-bearing body identification method fusing an unmanned aerial vehicle DOM and a residual Unet transfer learning algorithm, and the method comprises the steps: obtaining landslide data, controlling an unmanned aerial vehicle to carry out the simulated flight according to the landslide data, collecting image data, processing the image data, and obtaining a landslide orthoimage map; inputting the processed landslide orthoimage into a Res-Unet model to obtain a disaster-bearing body identification result; wherein the Res-Unet model is obtained by taking a pre-trained ResNet-34 network as a backbone network to replace an encoder structure of the Unet model, and updating parameters through training iteration by utilizing a training set; in the training process of the training set, the ResNet-34 network serves as a backbone network to carry out cross-layer transmission on training samples in the training set and add the training samples with a convolution result so as to consider the Res-Unet model with more features. According to the invention, quicker and more accurate target area monitoring and data acquisition can be realized, and the method is suitable for application scenes requiring high precision and high real-time requirements.

The invention belongs to a data processing method, and provides a data enhancement method, a data enhancement system, electronic equipment and a computer storage medium aiming at the technical problems that a manual annotation mode is difficult to cope with continuously changing business scenes and the timeliness and the flexibility of corresponding model training cannot be ensured when atlas construction and relation extraction technologies are utilized at present. And marking entities and relationships in the initial sample, performing classification arrangement to obtain an entity dictionary, replacing the entities of the text in the marked sample, and marking the replaced entities to obtain a first augmented sample. Constructing a relationship template and a sentence pattern template, re-expressing the text in the replaced first augmented sample through the sentence pattern template, replacing the entity and relationship of the text in the re-expressed first augmented sample in combination with the entity dictionary to obtain a second augmented sample, and then summarizing and fusing the first augmented sample and the second augmented sample to obtain a second augmented sample; and enhanced data is obtained.

The invention is applicable to the technical field of knowledge migration, and provides a graph domain adaptive node classification method based on instance learning, which comprises the following steps of: acquiring a source domain data set with a mark and a target domain data set without a mark; the representation of the source domain data set and the representation of the target domain data set are learned through a GNN model, modeling is conducted on the source domain data set and the target domain data set, and embedded vectors of the source domain data set and the target domain data set are obtained; the embedded vectors of the target domain data set are classified through KNN, and target domain category pseudo labels are obtained; performing fine-grained instance contrast learning on embedded vectors of the source domain graph data set and the target domain graph data set; and after classification is finished, displaying a final classification result of the target domain graph data set. According to the method, the defects of an existing graph domain adaptation technology in cross-domain tasks are overcome, and the method has wide applicability and can be applied to graph data analysis tasks in various cross-domain scenes, such as social network analysis, recommendation systems, knowledge graphs and biological networks.

The invention discloses a COD (Chemical Oxygen Demand) prediction method based on a two-channel convolution time sequence mechanism, which comprises the following steps of: acquiring COD original data for preprocessing to obtain COD input characteristics; constructing a prediction model, and importing data into a dual-channel convolution module; importing the first double-channel convolution output feature vector and the second double-channel convolution output feature vector into a double attention memory network to obtain a time sequence dependence feature; the time sequence dependence characteristics are imported into an output module, and predicted COD output is obtained; training the model, and in the training process, optimizing hyper-parameters of the model based on an improved pelican algorithm; and importing real-time COD original data into the trained model, and obtaining real-time COD output. According to the method, a high-efficiency effluent COD content prediction model of the sewage treatment plant is constructed by creatively fusing the two-channel convolution and the double-attention memory network.

The invention relates to a straight line and octagonal Steiner minimum tree construction unification method based on deep reinforcement learning, and belongs to the technical field of integrated circuit computer aided design. The method comprises the following steps of: firstly, designing an Edge Point Sequence (EPS) according to the structural characteristics of the SMT to be used for bridging the difference between the output of a deep learning model and the SMT structure; secondly, a deep learning model is designed for EPS, and the model is trained through DRL with the negative line length of SMT as an award; secondly, a corresponding rapid and accurate line length calculation algorithm is provided for quality evaluation of the construction scheme, so that model training is accelerated; and finally, utilizing the randomness of a machine learning method to construct diversified SMT construction schemes. According to the method, the RSMT and OSMT problems can be solved, and diversified wiring topologies can be generated.

The invention provides a three-dimensional Gaussian splash algorithm training method and device, equipment and a medium, and belongs to the technical field of image processing, and the method comprises the steps: obtaining an image data set, carrying out the down-sampling of an image in the image data set based on at least one sampling rate, and determining a target image data set; determining a three-dimensional Gaussian distribution model according to the image data set; performing multi-round iterative training on the three-dimensional Gaussian distribution model according to the target image data set and the two-dimensional filter; each round of iterative training determines an anti-aliasing filter according to the target image data set and a preset two-dimensional filter; and filtering the three-dimensional Gaussian distribution model according to an anti-aliasing filter, and training the filtered model. According to the method, iterative training is carried out by using a low-resolution image instead of an original image, the calculation complexity is reduced, training acceleration is realized, the three-dimensional Gaussian distribution model is filtered through the anti-aliasing filter, the aliasing effect is overcome, and the rendering quality is ensured while the training time is shortened.

The invention provides an informatization intelligent drainage management method and system, and relates to the technical field of drainage management, and the method comprises the steps: obtaining water level data, first flow data and pump station flow control data, and obtaining a water level change function; determining a predicted rainwater inflow amount, predicting a water inlet pipeline inflow amount and determining predicted water inlet flow data; and determining predicted pump station flow control data according to the predicted water inlet flow data and the water level change function, and adjusting the pump station of the drainage pipeline at the beginning of the next monitoring period. According to the method and the device, the water level change function capable of describing the relation among the water level data, the first flow data and the pump station flow control data can be determined, then the predicted pump station flow control data of the pump stations of the drainage pipelines of the water storage areas are determined, the drainage flow of the water storage areas is optimized, the drainage flow of the areas can be coordinated, and the drainage efficiency is improved. The water level of each water storage area is in a reasonable range, and risks such as overflow of the reservoir in each area are reduced.

The present invention relates to the field of image defogging, and specifically discloses a 4K image defogging method, device and storage medium based on weakly supervised contrast learning, including S1, obtaining a dark channel prior image #imgabs1# after dark channel prior processing of a foggy image #imgabs0#; weighted fusion of the foggy image #imgabs2# and the processed image #imgabs3#, and finally obtaining a fused image #imgabs4#; S2, inputting the fused image #imgabs5# into a U-net convolutional neural network structure for processing, and inserting a residual block and introducing an SC attention mechanism during processing; S3, using PatchGAN as a discriminator, which divides the restored image #imgabs6# into many small overlapping areas, judging whether each small area is part of the real image, using the least squares GAN loss to train the network, and finally discriminating and obtaining the restored fog-free image #imgabs7#; S4, contrast learning loss function. The beneficial effects achieved by the present invention are: the high-resolution image defogging algorithm is short in time and can maintain a high-precision defogging effect.

The invention provides a high-precision attitude estimation method for processing limited data based on a pre-training diffusion model, and the method comprises the steps: carrying out the preprocessing of an input image, and calling a function to combine with noise to generate a key point prompt; mapping the input image through an encoder to obtain a potential code, and increasing Gaussian noise of different amplitudes on the potential code to obtain a noise image; performing attention mechanism processing on the noise image to obtain a semantic corresponding point between the input image and the reference image; training the image pair, integrating training results of the basic class image and the new class image, and inputting an integration result into an Adam optimizer for optimization to obtain an optimization result; evaluating the image pair; according to the method, the key points of the unlabeled image data can be learned in a cross-category mode, a good attitude estimation effect is achieved under the condition that the data is limited, and the actual application value of attitude estimation is improved.