64 results about "Hybrid learning" patented technology

The invention relates to the technical field of agent path planning, in particular to a multi-agent collision-free path planning method based on a fusion DQN algorithm. The multi-agent collision-free path planning method comprises the following steps: firstly, constructing a two-dimensional grid map as an environment, and carrying out feature extraction by utilizing a CNN (Convolutional Neural Network); then, behavior clone learning is carried out through the expert model to obtain a BC model; the core innovation lies in that a BC model and a CNNDQN model are fused, an adaptive strategy learning framework is constructed, and intelligent dynamic combination of expert experience and reinforcement learning exploration is realized by adopting uncertainty estimation, antagonistic knowledge distillation and performance perception sampling technologies; and finally, further processing an initial path output by the fusion model by a CBS algorithm, and completing multi-agent collision-free path planning. According to the method, the accuracy and efficiency of path planning are optimized through a mixed learning strategy.

The present disclosure provides a computer-implemented system for automated augmentation of differential equation models. The system stores differential equations representing mechanistic behavior of physical or computational processes and constructs a hybrid computational solver by embedding a trainable universal approximator with adjustable parameters into the differential equations, where approximator outputs augment time derivatives during numerical integration. An iterative training process adjusts parameters through numerical integration, monitors integration failures, assigns infinite penalty values to loss functions when failures occur, and computes gradients using automatic differentiation otherwise. The system computes sensitivity metrics via Jacobian matrix evaluation, classifies input / output subsets as significant based on threshold-exceeding sensitivity metrics, generates a reduced approximator operating on classified subsets, and replaces the universal approximator with the reduced version to create an optimized solver.

The invention relates to the technical field of automatic driving and intelligent control, in particular to a multi-transportation-equipment cooperative motion control method based on multi-stage mixed learning and a storage medium, and the method comprises the steps: generating expert demonstration data through employing a single-vehicle motion control algorithm based on an expert rule, performing initialization training on the strategy network through online iterative supervised learning to obtain a pre-training strategy model; the multi-vehicle cooperative motion control method comprises the following steps: selecting a multi-vehicle cooperative motion model, loading parameters of the model into an Actor strategy network of multi-agent reinforcement learning, performing interactive training on a plurality of agents in a simulation environment by adopting a centralized training and decentralized execution normal form, and performing online iterative optimization on the strategy based on a composite reward function and generalized advantage estimation to obtain a multi-vehicle cooperative motion control strategy. According to the method, the complementary advantages of imitation learning and reinforcement learning are exerted, the training efficiency, the strategy performance and the collaborative operation capability and robustness of the system in a complex scene are improved, and the method can be directly applied to collaborative scheduling and control of transportation equipment groups in scenes such as surface mines and ports.

The invention discloses a multi-energy load prediction method based on feature screening and multi-model fusion, and belongs to the field of electric power and comprehensive energy load prediction. The invention provides a three-stage hybrid learning prediction framework. In the first stage, dynamic feature screening is achieved through a recursive feature elimination cross validation method based on expert knowledge constraints, key meteorological elements and multi-energy load time sequence features are reserved, and redundant feature interference is reduced. In the second stage, a multi-task long-short-term memory network is constructed, and coupling relation modeling and collaborative prediction of cold, heat and electricity multi-energy loads are achieved through sharing time sequence characteristic representation and a task exclusive output structure. And in the third stage, a random forest is adopted to carry out nonlinear correction on the residual error of the sub-model, so that the precision and robustness of prediction in sudden disturbance and local non-stationary scenes are improved, the prediction error is effectively reduced, and the stability of multi-energy load prediction is improved. And reliable support is provided for optimized operation, scheduling decision and renewable energy consumption of the park integrated energy system.

The invention relates to a key flow-based SRv6 explicit path intelligent scheduling method and system, and belongs to the technical field of communication information. The method comprises the following steps: carrying out criticality scoring through a semi-supervised-migration mixed learning model, and outputting a key stream; constructing a full-dimensional network twinborn model, pre-judging a congestion risk trend, and verifying a link path survival rate; generating a low-delay source node list through a key flow path enhancement algorithm based on the key flow and link path survival rate; analyzing the low-delay source node list to create a micro-isolation queue; when the digital twin detects that the actual state deviates from the predicted value, calling a pre-verified backup path identification chain to complete path switching in real time; and performing closed-loop optimization based on the key operation state data by using a double-channel learning model. The technical problems that in the prior art, key flow recognition precision is low, SRv6 path scheduling and network state adaptability is poor, fault switching lags behind, and core model optimization is insufficient are solved.

The invention relates to a non-intrusive load decomposition method and device based on a parallel mixed learning model, and the method comprises the steps: obtaining to-be-predicted load active power data, carrying out the preprocessing, inputting the data into a pre-trained parallel mixed deep learning model, and obtaining a load decomposition result; the parallel hybrid deep learning model comprises a CNN channel and a GRU channel which are arranged in parallel, and a feature splicing layer which is connected with the CNN channel and the GRU channel, the CNN channel is used for carrying out feature space dimensionality reduction on input features so as to obtain electric appliance space features, the GRU channel is used for extracting load time sequence dynamic features according to the input features, and the load time sequence dynamic features are matched with the feature splicing layer; and the feature splicing layer is used for fusing the electric appliance space features and the load time sequence dynamic features, and obtaining a load decomposition result through a meta-learner. Compared with the prior art, the method improves the efficiency of non-intrusive load decomposition, and has excellent performance and generalization ability in a complex power consumption scene.

The invention relates to the field of AI visual detection, in particular to an AI visual detection method based on supervised and unsupervised deep learning, which comprises the following steps: establishing a sampling image visual feature by using real-time sampling image data; establishing an AI visual feature analysis model based on supervised and unsupervised deep learning by using the visual features of the sampled image; according to the AI visual feature analysis model, an AI visual detection result is obtained, high-precision and high-robustness visual detection is achieved under a small amount of annotation data through a supervised and unsupervised mixed learning architecture in combination with the discrimination ability of deep learning and the anomaly detection advantage of unsupervised learning, a supervised main model is trained based on the annotation data, and a supervised auxiliary model is trained based on the unsupervised learning. Precise recognition of known defects is ensured, limitation of a traditional supervised model is broken through, generalization ability is improved while high precision is ensured, and the method is suitable for complex and changeable industrial detection scenes.

The invention provides a heat supply system thermal load prediction method based on time sequence compensation and hybrid learning, and the method comprises the steps: constructing a thermal load prediction input variable screening mechanism based on two-dimensional analysis, and achieving the scientific screening of input variables through the dual verification of a Pearson's correlation coefficient matrix and a significance test; a thermal load-temperature time sequence synchronization system based on dynamic time shift compensation is innovatively designed, and the problem of time sequence dislocation caused by instability of manual experience adjustment in a central heating system is solved; an ES-LSTM collaborative prediction architecture is provided, and a three-level prediction model of'trend decomposition-random learning-dynamic weighting 'is established. According to the method, statistics significance test and thermodynamic mechanism analysis are combined in a breakthrough manner, a dynamic compensation system with a time sequence self-correction capability is innovatively researched and developed, a hybrid model collaborative prediction mechanism is constructed, and a complete technical chain from data preprocessing to prediction model architecture is formed.

The invention discloses a data center global temperature control optimization method based on online-offline hybrid reinforcement learning, and the method comprises the steps: constructing an interpretable global state space fusing physical information, and guaranteeing that a state variable has a clear physical meaning; a combined action space is defined, and cooperative control of the cold source side and the tail end side is achieved; designing a multi-objective reward function based on physical model driving, and comprehensively considering energy consumption, temperature stability, carbon emission and physical optimization indexes; obtaining a basic security policy through offline pre-training, and extracting reliable behaviors from historical data by using a conservative Q learning algorithm; strategy optimization is achieved through online safety fine adjustment, and gradual adjustment is conducted under multiple constraints to adapt to real-time changes; and finally, deploying an optimization strategy to realize closed-loop control of the system, and establishing a continuous learning mechanism to cope with long-term drift. According to the method, a complete offline-online mixed learning system is established, and the energy efficiency of the data center temperature control system is improved.

The invention discloses a multi-domain combat dynamic deduction system based on deep reinforcement learning, and relates to the technical field of military automation, and the system comprises an environment perception-confrontation game-emergence mining three-layer nested dynamic architecture in which a bidirectional feedback mechanism, a cross activation mechanism and an adaptive adjustment node are deployed. The three-layer nested dynamic architecture comprises a dynamic adaptive adversarial network, a chaotic data weaving device and an emergence capture and evolution guider; according to the invention, through a three-layer nested architecture and core module design, multi-domain information real-time interaction and cross-level collaboration are realized, and the multi-domain linkage simulation precision is improved; by means of a heterogeneous subnet dual-mode interaction and mixed learning mechanism, the agent decision-making flexibility is enhanced; dynamic data are efficiently processed and deep association is mined by utilizing incremental topology analysis and a knowledge graph; and through anti-disturbance and propagation simulation, a chaos edge emergence phenomenon is captured, and finally, deduction and actual combat deviation is reduced through multi-module cooperation.

The invention belongs to the field of intelligent prostheses, and particularly relates to an intelligent prosthetic control algorithm based on EMG electromyographic signals, which comprises the following steps: S1, electromyographic signal acquisition: acquiring the electromyographic signals of the forearm of a user through a multi-channel surface electromyographic sensor, the electromyographic signals being sEMG signals; s2, signal preprocessing; according to the time-frequency feature enhancement and mixed learning architecture, the system can realize more accurate and stable small sample gesture recognition, the processing capability and classification performance of the non-stationary electromyographic signals are remarkably improved, and in addition, due to the fact that deep features and discriminative manual features are fused and a lightweight network and a multi-modal decision mechanism are combined, the recognition efficiency of the non-stationary electromyographic signals is improved. A user can realize reliable and smooth gesture interaction in a low-delay and low-data-dependence embedded environment, and does not need to depend on a large amount of labeled data or complex computing resources, so that the whole method has higher adaptability and practicability.

The invention belongs to the technical field of robots, and discloses a man-machine mutual teaching system, which comprises a humanoid robot, and is characterized in that the humanoid robot is provided with a multi-airbag bionic muscle system and an acupuncture point internal and external sensing network, is provided with a preset basic function module, is provided with a mixed learning framework, and is used for integrating end-to-end learning and multi-modal decomposition learning capabilities; the humanoid robot is further provided with a kappa coefficient library, a skill perception module, a ternary skill characterization module, a skill reproduction module, a skill optimization module, a mixed skill library, an isomorphic skill transfer module, a heterogeneous skill adaptation module, a staged skill transaction platform, a coevolution module and an AI intelligent optimization engine. The advantages of end-to-end learning and multi-mode decomposition learning are integrated through a mixed learning framework, the learning efficiency is improved, and the limitation of a single learning normal form is solved; the kappa coefficient library realizes systematic management of force control parameters, provides optimal kappa value configuration for different skills, and greatly improves motion flexibility and accuracy.

The embodiment of the invention discloses a transmission task scheduling optimization method and device based on a hybrid learning strategy, and relates to the field of the Internet of Things and artificial intelligence. The method comprises the following steps: constructing a task scheduling optimization model taking minimization of the total bandwidth cost of a system in a preset time period as a target function, initializing parameters of a hybrid learning strategy particle swarm optimization algorithm, including particle swarm scale, particle position and speed, iteratively optimizing a bandwidth allocation scheme according to the hybrid learning strategy particle swarm optimization algorithm, and optimizing the bandwidth allocation scheme according to the particle swarm optimization algorithm. And updating the individual historical optimal position of each particle and the global historical optimal position of the population in each iteration process, and when iteration reaches the maximum number of iterations, outputting a bandwidth allocation scheme corresponding to the global optimal position as an optimal transmission task scheduling scheme. In this way, fast and efficient allocation of network bandwidth resources is achieved.

The invention discloses a sparse social network-oriented graph mixed learning friend recommendation method, which comprises the following steps of: establishing an original graph structure data set, and respectively generating a diversified social network comparison view and a de-noising comparison view; performing comparative learning to obtain an initial feature embedding representation; updating initial feature embedding; fusing the updated feature embedding representation, and establishing a user node link probability prediction model; performing friend prediction, generating a plurality of user-user pseudo links, and obtaining a candidate pseudo link set; performing reliability screening to obtain a final pseudo-link set; obtaining a new graph structure data set; and replacing the original graph structure data set with the new graph structure data set, and repeating the steps until a new user-user pseudo link cannot be generated. The method is used for solving the problems that in the prior art, potential friend relations are difficult to fully excavate in the face of sparse social networks, and recommendation results are not accurate, and the purpose of better adapting to friend recommendation requirements in the sparse social network environment is achieved.

The invention discloses a hybrid learning-based flexible job shop energy-saving batch scheduling method, which relates to the technical field of shop scheduling and comprises the following steps of: establishing a flexible job shop energy-saving scheduling model by taking minimization of maximum completion time and total energy consumption of a machine as optimization objectives; the method comprises the following steps of: constructing a mapping relationship between an operation process and machine selection by adopting a three-layer coding mode, initializing a population through multiple initialization strategies, decoding a coding part, and generating an initial solution; constructing a neighborhood structure, performing population evolution through an adaptive crossover and mutation operator, and adjusting the crossover and mutation probability according to the population evolution degree; searching an optimal solution through a plurality of search strategies; traversing the population to perform non-dominated sorting, and adjusting the crossover mutation rate according to a non-dominated sorting result; and judging whether an iteration termination condition is met or not, if not, continuing iteration, and if so, outputting the optimal Pareto frontier solution. According to the invention, the energy consumption is reduced while the processing time is minimized.

The invention discloses a controller fault diagnosis method and system, a computing device and a readable storage medium. The method comprises the following steps: synchronously acquiring a multi-mode signal in the operation process of a controller; fusing the multi-modal signals to obtain a fused feature vector; and inputting the fusion feature vector into a hybrid learning model to obtain a fault diagnosis result of the controller. Therefore, by synchronously collecting and fusing the multi-mode signals and utilizing the mixed learning model to diagnose and break the fault, the fault detection precision and robustness are remarkably improved, and high-sensitivity and low-delay diagnosis under the complex working condition is achieved.

A kind of industrial robot dynamics and friction reconstruction method based on PINN, in offline stage, robot excitation trajectory is constructed and robot excitation experiment is set, and joint position data, speed data and current data of robot in process are collected to construct data set, PINN model of robot dynamics modeling based on Lagrange method and PINN model of robot friction modeling based on coulomb viscous model are mixed learning;In online stage, the trained robot dynamics PINN model and robot friction PINN model are used for real-time inference prediction, and the dynamics component and friction component of the joint torque of the robot are obtained.The present application significantly simplifies the modeling and prediction process of the joint torque of the industrial robot, and can realize high-precision joint torque prediction effect, thereby improving the overall motion control performance of the industrial robot based on the dynamics model.

A communication networking method applicable to wide-area decentralized charging piles includes: initializing the total number of wide-area decentralized charging piles and coordinate parameters thereof; initializing related parameters of a hybrid learning network, and predicting and judging optimal networking nodes of the charging piles by using the hybrid learning network; and performing connection according to the optimal networking nodes, so as to form a communication network of the wide-area decentralized charging piles. With regard to the communication networking applicable to the wide-area decentralized charging piles, the communication networking method evaluates a channel environment between the wide-area decentralized charging piles, and screens out optimal master nodes according to the quality of a communication environment between the charging piles, so as to form a communication networking scheme of the wide-area decentralized charging piles. In this way, when the distances between the charging piles are relatively large, the communication reliability can be guaranteed.

The invention discloses a novel power system false data injection attack detection method based on CPO-CNN-SE. The method comprises the following steps: preprocessing a measurement test of a novel power system; performing hyper-parameter optimization by using a crown porcupine optimization algorithm (CPO) to improve the precision of the CNN detection model; meanwhile, in consideration of synchronous increase of the FDIA detection speed, a compression excitation (SE) attention mechanism is introduced, a CNN-SE mixed learning model is formed, and the characterization efficiency of the CNN on FDIA spatial features is enhanced; and finally, carrying out binary classification on the FDIA feature data through a full connection layer, a sigmoid classifier and a binary cross entropy loss function. According to the invention, the FDIA detection precision is improved, and the synchronous improvement of the detection speed is realized.

The application discloses an unbalanced target classification method, product, medium and equipment, relates to the field of class unbalanced learning of image classification, and solves the problem of poor classification effect caused by class unbalanced learning in real domain data through mixed learning of real domain data and virtual domain data, that is, through virtual-real data mixed training, improves the classification accuracy of tail classes in the real domain, thereby improving the target classification performance in the real domain data. Meanwhile, the application solves the domain offset problem through hidden layer feature enhancement in the network, improves the target classification precision in the real domain, and finally achieves the effect of greatly improving the offshore target classification accuracy in the real domain.

The application discloses a kind of outbound line decision-making method and system based on incoming call display perception, the method includes: determining outbound task and obtaining the number attribute of outbound line, the number attribute and outbound task are modeled to obtain number feature vector and outbound service feature vector;Based on the willingness prediction network of pre-set, according to number feature vector and outbound service feature vector, the willingness score sequence corresponding to outbound line is obtained by prediction;For any outbound line, according to the constraint condition of willingness score and line cost, the target outbound line executed by outbound task is determined by calculating through the pre-set resource allocation algorithm;The feedback sample of target outbound line corresponding outbound task is obtained, and the willingness prediction network and resource allocation algorithm are continuously updated by hybrid learning mechanism.It can be seen that the application can realize the intelligent selection and dynamic optimization of outbound line, so as to effectively improve the reach rate of intelligent outbound system.

A computer system and method for training a machine learning system to perform a classification task by dividing input data into one of a plurality of classes. The system is configured to receive training data for each class from which a representation of each class can be derived, wherein each class is described by a plurality of representations; process the training data to form, for at least one class, a first proxy for a relatively global portion of training data items and a plurality of proxies for different relatively local portions of training data items, each proxy corresponding to a representation of data belonging to the class. For each training data item, the system is configured to: evaluate a match between the training data item and the proxies; estimate the class of the training data item from the match rating; adjust the proxies by updating a weighting matrix to reduce the distance between the training data item and the proxies of the estimated class. Defining a plurality of proxies in this way can result in richer and more stable representations of object classes.

The application provides a hybrid learning type image compressive sensing method and system based on semantic guidance, belongs to the technical field of image processing and compressive sensing, and comprises the following steps: acquiring an input image, dividing the input image into a plurality of non-overlapping image blocks, inputting the image blocks into a semantic coding network, and using the semantic coding network to extract high-dimensional semantic embedding features of the image blocks and generate corresponding semantic mask vectors; constructing a semantic guidance adaptive sampling matrix based on the semantic mask vectors, so that the sampling matrix dynamically changes with the semantic features of the image blocks; after obtaining the adaptive sampling matrix, performing a compressive measurement operation on the image blocks to obtain basic sampling measurement values and semantic adaptive sampling measurement values, respectively, splicing the two measurement results in the measurement dimension to obtain a final compressive measurement vector; and sequentially performing semantic perception pre-reconstruction, hybrid Transformer-CNN block level reconstruction and pixel level fine reconstruction based on the compressive measurement vector, and outputting a reconstruction result.

A hybrid learning method for intracranial electroencephalogram (EEG) signal processing based on contrastive learning and mask reconstruction is proposed. After acquiring EEG signals offline and constructing a training set, an IntraBraM network is built, comprising a convolutional network block encoder, a Transformer encoder, a decoder, and a linear prediction layer. Following contrastive and mask reconstruction training, the trained IntraBraM is used for real-time EEG signal classification online. This invention addresses both the pre-training paradigm and model structure design, creating a pre-training paradigm and model structure adapted to the intracranial EEG modality. This enhances the model's performance on downstream tasks within the intracranial EEG modality, provides robustness to the spatial coordinates of adversarial electrodes, and improves generalization ability across subject settings.

The invention belongs to a cell recognition technology, and particularly provides a cell focusing and detection method based on mixed learning, which comprises the following steps: A1, performing cell recognition in an initially focused image by using a depth target detection algorithm, and deleting bounding boxes of which the area is greater than a threshold ST to obtain bounding boxes of suspected cells; a2, performing automatic focusing on the cell surface by using a DCS focusing method; a3, performing information enhancement on the image; and A4, inputting the enhanced cell picture into a YOLO-DeCDIT (Yolo-DeCDIT) algorithm to carry out cell recognition and positioning, wherein the C2PSA module in the yolk is replaced by DeCDIT by the algorithm. The method has the advantages of high cell recognition accuracy and the like.

The present invention discloses a method and device for processing electric power images based on complete learning. The method uses a feature map Fout and a calibration value of the feature map Fout as a training set, and uses the training set to detect anomalies in the electric power image. The electric power image is input into the trained electric power image anomaly detection model, and the prediction result of anomalies of electric power equipment in the electric power image is output. The advantages of local feature extraction of the convolution process and the advantages of global feature extraction of the self-attention calculation are effectively utilized to construct an efficient image feature learning method. The feature map of hybrid learning can fully represent the feature information of the original image and effectively learn the feature information in the electric power image. The present invention improves the accuracy of recognition, reduces the cost of manual inspections, automatically detects defects on transmission lines, and ensures the safe operation of the national power system.

The embodiment of the application provides a wind turbine full life cycle prediction method and device based on multi-physical field coupling, which comprises the following steps: dividing the whole structure of the wind turbine into structural components and related spaces, collecting multi-physical field sensor data from the whole structure, discretizing the structural components of the wind turbine into a plurality of finite elements, defining the related spaces as nodes connected to the finite elements, constructing a corresponding wind turbine network topology, establishing corresponding physical field control equations for the finite elements based on the network topology, determining a corresponding global coupling equation set, and solving the global coupling equation set by using a numerical method to determine the corresponding physical response data. Real-time multi-physical field sensor data and real-time physical response data are input into a set hybrid learning model for unsupervised and supervised hybrid learning to determine the life cycle prediction result of the wind turbine. The application can improve the efficiency and accuracy of wind turbine fault prediction.

Introduced here is an interactive digital health system (or simply “interactive system”) that is able to update a knowledge base used to facilitate semi-automated conversations to account for feedback. The interactive system can include a collection of modules that, in operation, can provide or support coaching services in the digital health domain, Upon receiving input indicative of a question from a patient, these modules allow the interactive system to perform language modeling to identify content objects in a knowledge base that are relevant to the question. Moreover, the interactive system may update the knowledge base to account for feedback, thereby making the knowledge base more robust to a broad range of questions.

An object of the system according to the embodiment is to increase the time for questions and answers in an online or hybrid study meeting or lecture meeting.SOLUTION: A system includes a chat insertion part, a reaction analysis part, and a question generation part. The chat insertion unit automatically inserts a chat. The reaction analysis unit analyzes a reaction of the participant to the chat inserted by the chat insertion unit. The question generation unit automatically generates a question based on the reaction analyzed by the reaction analysis unit.SELECTED DRAWING: Figure 1

The invention provides a method for establishing a value-oriented content selection model in an ideological and political education plan, and belongs to the technical field of neural network models. A memory matrix is realized in a plurality of hidden layers to store historical semantic vectors, core features are selectively reserved through a forgetting gating unit, a symbol knowledge base is established to store value logic rules and convert the value logic rules into differentiable soft logic constraints, and a sparse attention mechanism and a gradient check point technology are adopted to reduce calculation complexity. And setting an adaptive gradient cutting threshold value and a dynamic learning rate zoom factor to perform model training, thereby solving the technical problem that effective modeling of the long text cross-paragraph value logic consistency in the ideological and political education plan value oriented evaluation process is difficult.