Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

84 results about "Selection strategy" patented technology

Selection strategies are the result of many design decisions, and it is safe to say that no two strategies are the same. The purpose of this entry is to describe different selection strategies and evaluate the effectiveness of those strategies in different employment situations.

Test case determination method and device, computer readable storage medium and processor

The invention discloses a test case determination method and device, a computer readable storage medium and a processor. The method comprises the following steps: identifying a source code of software by utilizing a pre-trained large model to obtain a plurality of functional modules in the software and a dependency relationship among the plurality of functional modules; in response to updating of the source code of the software, determining a code updating part in the source code; determining a test case matched with the code updating part from a test case library of the software based on the code updating part and a dependency relationship among the plurality of functional modules, the test case being used for testing the code updating part; based on a test result of the test case, adjusting a selection strategy of the test case; and re-determining the test case for testing the code updating part based on the adjusted selection strategy of the test case. The technical problem that a traditional regression test strategy is low in test efficiency is solved.
Owner:CHINA TELECOM ARTIFICIAL INTELLIGENCE TECHNOLOGY (BEIJING) CO LTD

Multi-agent-based content recommendation method and device, equipment, medium and product

The invention discloses a multi-agent-based content recommendation method and device, equipment, a medium and a product, and the method comprises the steps: inputting obtained user data into each pre-constructed recall agent, and obtaining recommendation content outputted by each recall agent; inputting the user data into a pre-constructed large language model to obtain a selection strategy of each recall agent; and according to the selection strategy, carrying out fusion processing on the recommendation content output by each recall agent to obtain fused recommendation content. According to the scheme, through fusion of the recommended contents of the multiple recall agents, a more comprehensive recommended content output scheme is provided, the scene application range is widened, content fusion is carried out based on the calculated selection strategy, and the content recommendation accuracy is improved.
Owner:MIGU CO LTD +1

Continuous strategic alignment in cloud-based environments

An embodiment predicts a plurality of strategies associated with a cloud workload. The embodiment presents the predicted plurality of strategies to a user for selection of a strategy. The embodiment predicts a first technology for the cloud workload having a first technology strategic potential value based on the selected strategy. The embodiment predicts a first deployment for the cloud workload having a first deployment strategic potential value based on the selected strategy. The embodiment identifies a second technology for the cloud workload having a second technology strategic potential value based on the selected strategy. The embodiment identifies a second deployment for the cloud workload having a second deployment strategic potential value based on the selected strategy. The embodiment transmits a strategic alert.
Owner:INTERNATIONAL BUSINESS MACHINE CORPORATION

Integrated machine learning model pre-labeling method and system

The invention discloses an integrated machine learning model pre-labeling method and system. The method comprises the following steps: S1) importing data; s2) model selection and matching; s202) multi-model selection and automatic optimization: comprehensively selecting an optimal model based on a selection strategy of data features and task complexity; s3) automatic configuration of the model; S4) pre-labeling of the model; s5) performing manual auditing and correction; and S6) outputting a labeling result. S7) performing model performance evaluation, and automatically adjusting hyper-parameters; s8) updating a model recommendation strategy through a random forest algorithm; the system comprises a user module, a data import module, an annotation type selection module, a model configuration module, an annotation module, an auditing and adjusting module, an annotation result output module and a model adjusting module. According to the method, the machine learning model is integrated to realize pre-annotation of the data, meanwhile, the recommendation algorithm based on the annotation type is designed, the machine learning model is recommended according to the annotation type selected by the user, and the annotation accuracy and speed are improved.
Owner:KUNSHAN XINGTUQIHANG ARTIFICIAL INTELLIGENCE TECHNOLOGY CO LTD

Traveling salesman problem solving method combining deep reinforcement learning and heuristic algorithm

PendingCN119106778AReduce total transportation distancelow costBiological modelsLearning basedData set
The invention provides a deep reinforcement learning and heuristic algorithm combined traveling salesman problem solving method. The method comprises the steps of defining a task of learning an improved heuristic algorithm to solve a traveling salesman problem as a Markov decision process; designing a new model M based on Transform, parameterizing a strategy network, and introducing an action discarding mechanism to prevent action selection from over-fitting; a deep reinforcement learning method (RL-SA) combined with a simulated annealing mechanism is provided to learn a selection strategy of a node to an operator so as to improve the performance of a 2-opt algorithm; the RL-SA uses a whale optimization algorithm to generate an initial solution to improve the sampling efficiency, and uses a Gaussian perturbation strategy to solve the sparse reward problem in reinforcement learning. The method is obviously superior to an existing learning-based method in a TSPLIB data set and a real container liner shipping data set, and the gap between the method and a highly optimized solver is further reduced.
Owner:ZHEJIANG SIGANG LINKAGE DEV CO LTD +1

Low-carbon economic operation optimization method for electricity-hydrogen-heat comprehensive energy system

The invention discloses a low-carbon economic operation optimization method for an electricity-hydrogen-heat comprehensive energy system, and belongs to the crossing field of intelligent energy systems and artificial intelligence, and the method comprises the following steps: 1, building a low-carbon economic dual-target operation optimization problem model of the electricity-hydrogen-heat comprehensive energy system; 2, constructing a dual-objective operation optimization problem solving framework with a third-generation non-dominated sorting genetic algorithm as a core; 3, modeling a population evolution decision problem in the genetic algorithm into a Markov decision process, and constructing a deep reinforcement learning agent and population evolution environment interaction mechanism; 4, training the intelligent agent by adopting a double-depth Q network algorithm, and outputting a genetic action according to an environment state; 5, obtaining a new filial generation in combination with a selection strategy associated with the reference points and genetic actions output by the intelligent agent; the step 4 and the step 5 are iteratively evolved until the maximum iteration generation number is reached, and the operation cost can be reduced under the same carbon emission condition.
Owner:NANJING UNIV OF POSTS & TELECOMM

Electric shovel fault diagnosis method based on sparse evolutionary algorithm and neural network

The invention discloses an electric shovel fault diagnosis method based on a sparse evolutionary algorithm and a neural network. The method comprises the steps of 1, acquiring electric shovel data; 2, establishing a neural network; 3, generating a guide vector and selecting a parent by using a mating selection strategy in SPEA2; 4, clustering the decision variables into groups, and calculating the importance of each group according to the guide vector of the current population; 5, generating filial generations by using crossover mutation operators; and step 6, performing continuous iteration, selecting N solutions by using an environment selection strategy of SPEA2, and finally obtaining a group of solutions which meet the Pareto leading edge surface and have relatively high classification precision and relatively low neural network sparsity. Multiple solutions can be obtained through one-time operation to meet different electric shovel fault diagnosis requirements, and therefore the electric shovel fault diagnosis efficiency and precision can be improved.
Owner:SHANXI TZCO INTELLIGENT MINING EQUIPMENT TECHNOLOGY CO LTD

Methods for testing a computer program

According to various embodiments, a method for testing a computer program is described, comprising performing several test iterations, wherein for each test iteration a reference test input and a mutation for the reference test input are selected according to a selection strategy (201), and a test input for the computer program is generated by applying the selected mutation to the selected reference test input, wherein for each test iteration a test run of the computer program is performed with the respective generated test input for the computer program, wherein for each test iteration a test coverage increase (209) is determined which is achieved by the test run, and wherein the selection strategy (201) is trained by means of reinforcement learning, wherein the test coverage increases (209) are used as rewards (204).
Owner:ROBERT BOSCH GMBH

Point cloud denoising method and device, medium and product

The embodiment of the invention provides a point cloud denoising method and device, a medium and a product, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring target point cloud data subjected to abnormal point elimination; according to the target point cloud data, constructing a density-based hierarchical clustering algorithm and a de-noising network of an operation selection strategy, and generating a de-noising model based on the de-noising network; and inputting the to-be-denoised point cloud data into the denoising model, and outputting the denoised point cloud data. According to the scheme, isolated noise possibly misleading path decision is filtered in advance, and interference is cleared for follow-up path selection; a denoising network based on a density hierarchical clustering algorithm and an operation selection strategy is constructed, a point cloud structure is accurately divided, representative elite points are screened in combination with the operation selection strategy, under extreme conditions, the elite points are preferentially used as core extension paths, noise point dominant decision making is avoided, and the accuracy of the system is improved. The problem that in the prior art, a single path is poor in adaptability in an extreme scene is solved, and the denoising accuracy is improved.
Owner:CHINA MOBILE ZIJIN INNOVATION INST CO LTD +2

Hierarchical solving method for virtual maintenance attitude of aircraft in narrow space

The invention discloses a hierarchical solving method for a virtual maintenance attitude of an airplane in a narrow space, and the method comprises the steps: deeply analyzing a plurality of man-machine engineering factors which may affect an operation effect in the maintenance attitude, such as the movable range of each joint of a human body, the balance of the attitude, the comfort, the accessibility and the visibility; a multi-objective optimization model of a virtual maintenance posture is established for the upper body and the lower body for the complex factors. In consideration of special requirements of maintenance tasks in a narrow space, an optimization scheme of an upper body and a lower body is specially designed, so that the optimal operation effect can be achieved with the minimum body burden when maintenance personnel carry out maintenance operation, and secondly, an NSGA-II (non-dominated sorting genetic algorithm II) algorithm for improving a selection strategy is introduced. According to the algorithm, through an improved selection mechanism, an optimal solution can be more quickly converged in a multi-objective optimization process, so that the generation speed and accuracy of the maintenance attitude are remarkably improved.
Owner:CIVIL AVIATION UNIV OF CHINA

Efficient and energy-saving flexible job shop scheduling method considering adjustment time

The invention relates to the technical field of intelligent manufacturing, in particular to an efficient and energy-saving flexible job shop scheduling method considering adjustment time. Comprising the steps of 1, initializing an initial population of an MTCP-SPEA algorithm; 2, executing a spatial awareness environment selection strategy; step 3, judging whether the current running time of the MTCP-SPEA algorithm reaches half of the total running time, if so, executing step 6, and otherwise, executing step 4; 4, a crossover operator and a mutation operator are adopted to evolve the current population, and current non-dominated individuals are selected to form a Pareto solution set; 5, executing local search operation and returning to the step 2; and step 6, executing a multi-thread-based constrained programming auxiliary optimization method to generate an improved elite set, screening out all non-dominated individuals from the improved elite set to form an optimal solution set, and outputting the optimal solution set. According to the method, the problem of efficient and energy-saving flexible job shop scheduling optimization considering time adjustment is effectively solved.
Owner:LIAOCHENG UNIV

LLM-based position description generation method and apparatus, and electronic device

The invention discloses an LLM-based position description generation method and apparatus, and an electronic device. The method comprises the steps of obtaining recruitment demand information input by a recruiter; obtaining enterprise portrait attributes of recruiters to obtain an enterprise information set and a function label set; the enterprise information set comprises multiple enterprise information labels and label values under the enterprise information labels; the function label set comprises multiple function labels and label values under the function labels; performing function estimation according to the recruitment demand information to obtain a first function corresponding to the to-be-recruitment post; constructing a position description candidate sample set according to the first function and the enterprise information set; according to the position description candidate sample set and the selection strategy, obtaining a position description target sample; on the basis of a preset prompt template, constructing a position description prompt according to the recruitment demand information, the position description target sample, the function label set and the enterprise information set; and inputting the constructed position description prompt into the LLM model to generate the position description.
Owner:QIAN JIN NETWORK INFORMATION TECH SHANGHAI LTD

Multi-AGV flexible job shop active scheduling method based on improved VNS-INSGA-II algorithm and related device

PendingCN120540231AForecastingKnowledge based modelsTournament selectionJob shop
The invention provides a multi-AGV flexible job shop active scheduling method based on an improved VNS-INSGA-II algorithm and a related device, and belongs to the technical field of shop active scheduling. According to the method, a multi-objective optimization model for active scheduling of the multi-AGV flexible job shop is constructed, and constraint conditions of the multi-objective optimization model are set; the VNS-INSGA-II algorithm is improved by adopting a multi-layer coding mode, a hybrid population initialization mode, a time-varying coefficient-based dual-strategy binary tournament selection strategy and a variable neighborhood search algorithm, and the improved VNS-INSGA-II algorithm is obtained; based on a set constraint condition of the multi-objective optimization model, an improved VNS-INSGA-II algorithm is adopted to solve decision variables in an objective function of the established multi-objective optimization model for active scheduling of the multi-AGV flexible job shop, and a Pareto frontier is obtained; and selecting a group of solutions from the Pareto frontier to obtain an active scheduling result, thereby carrying out active scheduling on the multi-AGV flexible job shop. According to the invention, the problems of low stability and low scheduling precision of the scheduling system are solved.
Owner:SHAANXI UNIV OF SCI & TECH

Project group reconstruction method and system for dynamic strategic target

The invention belongs to the field of project group decision making, and particularly relates to a dynamic strategic target-oriented project group reconstruction method and system, and the method comprises the steps: obtaining a project adjustment decision scheme according to a constraint condition; inputting the strategic targets, the project attributes and the decision-making schemes into a large language model to obtain preference scores of the decision-making schemes, and then sorting the decision-making schemes to obtain a sorting result; setting a state; the strategy network generates a corresponding weight vector according to the sorting result and selects an execution action to obtain a selection strategy; according to the initial state and the execution action, state conversion is carried out to obtain a next state, and the extraction weight of each training sample is calculated; updating the strategy network according to the extraction weight and the training sample to obtain an optimized strategy network; and obtaining a project group adjustment result according to the sorting result and the optimization strategy network. The method has the effects that the project is objectively decided, and the project combination is accurately changed according to the changing dynamic strategy target.
Owner:NAT UNIV OF DEFENSE TECH

Large model prompt thinking chain construction method for electric power customer service and related device

The invention belongs to the crossing field of artificial intelligence and an electric power system, and discloses an electric power customer service oriented large model prompt thinking chain construction method and related device.Firstly, a target topic type is determined according to an electric power customer service problem, and then comprehensive representation of all fused thinking chains of the target topic type is obtained; and selecting a fusion thinking chain corresponding to the comprehensive representation with the highest similarity based on the similarity, and taking the fusion thinking chain as a large model prompt thinking chain of the electric power customer service problem for the electric power customer service. According to the method, the selection strategy based on the topic type is adopted, the complexity of the selection strategy is effectively reduced, and the selection efficiency and the cue word quality are improved. Meanwhile, the comprehensive representation of each fusion thinking chain of the target topic type adopts a multi-thinking chain example fusion analysis strategy based on a graph attention network, and the reasoning process information of multiple thinking chain examples is integrated, so that the reasoning stability and accuracy of the finally selected fusion thinking chain are improved; the problem of instability caused by reasoning based on a single thinking chain example is effectively solved.
Owner:CHINA ELECTRIC POWER RESEARCH INSTITUTE CO LTD +3

Industrial equipment unsupervised abnormal sound signal detection method based on similarity feature selection decoupling

The invention discloses an industrial equipment unsupervised abnormal sound signal detection method based on similar feature selection decoupling, and the method comprises the steps: 1) carrying out the feature extraction of an original sound signal, and obtaining a logarithmic Mel spectrogram (feature matrix); 2) performing time domain feature selection on the logarithmic Mel spectrogram by using a similarity measurement-based method, and calculating the similarity between each frame in the spectrogram and the whole spectrogram; 3) according to the similarity in the step 2), designing different selection strategies by utilizing the characteristics of stable and non-stable sound signals; 4) processing sound signals of normal work of the industrial equipment according to the steps 1)-3) to obtain a data set after feature selection; 5) inputting the training set into a beta-variational auto-encoder, and training by taking minimization of a reconstruction error and KL divergence as optimization targets; 6) taking the mean square error as an abnormal score, and setting a threshold value according to the abnormal score of the training set; and 7) inputting the test set into the trained network to calculate an abnormal score, and judging whether the network is abnormal or not according to a threshold value.
Owner:ZHEJIANG SCI-TECH UNIV

Prediction Method, Encoder, Decoder and Storage Medium for Point Cloud Attributes

The present application provides a method for predicting point cloud attributes, an encoder, a decoder, and a storage medium. During the prediction process of point cloud attributes, different target adjacent point selection strategies are designed according to the distribution of duplicate points to determine at least one target adjacent point of the current point, and the attributes of the current point are predicted based on the reconstructed attribute information of the at least one target adjacent point, thereby improving the efficiency of point cloud attribute prediction.
Owner:TENCENT TECHNOLOGY (SHENZHEN) CO LTD

Strategy optimization method and device for multi-modal action model, equipment and medium

The invention relates to the technical field of artificial intelligence, can be applied to business system platforms of financial science and technology, medical treatment and health and the like, and discloses a strategy optimization method, device, equipment and medium of a multi-modal action model. Obtaining relation dependency, constructing an initial multi-modal action model in combination with the initial training parameter group, performing fine tuning on the initial multi-modal action model by using the obtained task specific data to obtain a fine-tuned multi-modal action model, and obtaining an environment interaction data set of a target environment; and performing interactive sampling on the environment interaction data set one by one by using the fine-tuning multi-modal action model to generate a plurality of target interaction tracks, and performing strategy optimization on a selection strategy in the fine-tuning multi-modal action model to obtain a target selection strategy. According to the method, the model selection strategy accuracy of the multi-modal action model is improved when the multi-modal action model faces a new situation or data is insufficient.
Owner:PING AN TECH (BEIJING) CO LTD

Multi-objective welding beam design method based on grey prediction evolutionary algorithm

The invention relates to the technical field of welded beams, in particular to a multi-target welded beam design method based on a grey prediction evolutionary algorithm, which comprises the following steps: randomly generating a first-generation solution set, a second-generation solution set and a third-generation solution set of a multi-target welded beam mathematical model; mining evolutionary trends of three solutions randomly selected from the first-generation solution set, the second-generation solution set and the third-generation solution set by using a grey prediction propagation operator, and predicting a solution in the intermediate-generation solution set; after the intermediate generation solution set is generated, constructing a candidate solution set, and performing non-dominated sorting on the candidate solution set to obtain a non-dominated leading edge set; selecting a fourth-generation solution set from the non-dominated leading edge set by using an adaptive environment selection mechanism, namely, autonomously switching a selection strategy according to the evolution condition of the solution; and finally, through algorithm iteration, a group of Pareto optimal solutions which are fully converged and are diverse and uniform in distribution are obtained. According to the method, the flexibility and adaptability of designing the multi-target welding beam are improved, and the application effect and economic benefits of the welding beam in actual engineering are promoted.
Owner:JIANGNAN UNIV

Memory management method and storage device

The present application provides a memory management method and a storage device. The method comprises: based on an operation event, obtaining a selection strategy corresponding to the operation event through a performance mapping table; determining a target virtual block combination from a plurality of candidate virtual block combinations according to the selection strategy; and responding to the operation event using the target virtual block combination. Thus, the performance and / or operation stability of the storage device can be effectively improved.
Owner:HEFEI KAIMENG TECHNOLOGY CO LTD

Distributed machine learning training configuration optimization method based on improved genetic algorithm

The invention belongs to the technical field of distributed machine learning training and configurable software system performance optimization, and particularly relates to a distributed machine learning training configuration optimization method based on an improved genetic algorithm, and the method comprises the following steps: S1, importing a performance prediction model; s2, feature gene coding; s3, performing population initialization; s4, evaluating the fitness; s5, selecting a strategy; s6, performing a cross strategy; s7, performing a self-adaptive variation strategy; s8, population replacement and integration; and S9, executing S4 and checking a termination condition, and if not, executing S5 to S8. The overall structure, the round training time and the video memory occupancy performance prediction model provided by the embodiment of the invention are optimized and can be popularized to distributed machine learning training systems with different architectures and network topologies, and meanwhile, priori knowledge in the distributed training process is fully utilized; the parallel configuration parameters are guided to search in the directions of shortening of round training time, increasing of video memory occupation and relatively balanced load division, and better optimization effect and convergence speed are achieved.
Owner:HEGANG XIONGAN DIGITAL TECH CO LTD +1

Point cloud attribute prediction method, encoder, decoder and storage medium

The invention provides a point cloud attribute prediction method, an encoder, a decoder and a storage medium, in a point cloud attribute prediction process, different target adjacent point selection strategies are designed according to the distribution condition of repeated points, at least one target adjacent point of a current point is determined, and according to the reconstruction attribute information of the at least one target adjacent point, the current point is predicted according to the reconstruction attribute information of the at least one target adjacent point. And attribute prediction is carried out on the current point, so that the point cloud attribute prediction efficiency is improved.
Owner:TENCENT TECHNOLOGY (SHENZHEN) CO LTD

Full-coverage path planning method

The invention discloses a full-coverage path planning method. The method comprises the following steps: determining a grid map of an area to be subjected to path planning, and determining a transition probability matrix corresponding to the grid map; an initial population corresponding to the grid map is determined according to the transition probability matrix and preset selection strategies, and the preset selection strategies comprise a greedy selection strategy and a fuzzy selection strategy; determining a local distance contribution value of a target path in the initial population, executing a multi-modal disturbance operation on the target path according to the local distance contribution value, and generating a filial generation population corresponding to the initial population; and performing iterative optimization according to the initial population and the offspring population until an optimal path corresponding to the grid map is determined. According to the invention, technical problems of path redundancy, slow convergence speed and poor adaptability to a complex environment existing in a path planning algorithm in a complex obstacle scene in the prior art are solved.
Owner:CHINA TELECOM CORP LTD

Multi-objective optimization design method and system for automobile top compression resistance structure considering parameter uncertainty

The invention discloses a multi-objective optimization design method and system for an automobile top compression resistance structure considering parameter uncertainty, and aims to solve the problem that reliability, safety and light weight are difficult to consider into consideration due to performance fluctuation of a design scheme under the influence of the parameter uncertainty. The method comprises the steps that parameter uncertainty is represented in an interval form, and an interval multi-objective optimization model of the top compression resistance structure is established; after the design group and the archiving set are initialized, a multi-output Gaussian process model is established for each optimization target so as to provide accurate response prediction; executing a dominating-decomposing dual-mechanism search strategy to guide the design group to evolve, and improving the quality of an optimal solution; and outputting the optimal design scheme in the archiving set through an interval inflection point selection strategy. Compared with a traditional optimization design method, the solution set reliability comprehensive index is improved by 11% +, the design efficiency is improved by 50.0%, the requirements for reliability, safety and light weight are effectively considered, and the design period is greatly shortened.
Owner:CENT SOUTH UNIV

Active domain adaptation semantic segmentation method, system, device and storage medium

The application discloses an active domain adaptation semantic segmentation method, system, device and storage medium, selects and labels taking superpixels as units, which is different from image-level and pixel-level labeling methods, and the superpixel-level labeling greatly improves labeling efficiency by only assigning a semantic category to each superpixel; in addition, different from the existing scheme based on the uncertainty selection labeling strategy, the application focuses on difficult example samples in the domain adaptation scene, and proposes a selection strategy based on domain information quantity to label superpixels most valuable for domain adaptation learning; by adopting the superpixel-level labeling method and the selection strategy based on the domain information quantity, the application greatly reduces the labeling cost while improving the labeling quality, and guarantees the performance of the domain adaptation semantic segmentation.
Owner:UNIV OF SCI & TECH OF CHINA

Ship transport capacity configuration optimization decision-making method and system, medium and server

The invention belongs to the technical field of transport capacity scheduling, and provides a ship transport capacity configuration optimization decision-making method and system, a medium and a server, and the method comprises the steps: predicting the transport price of each route in each plan period in the future through employing Neuro-ARIMA based on the historical transport price data of energy and bulk material waterway transportation; constructing a transport capacity configuration optimization model by taking net income maximization in a planning period as a target; and solving the transport capacity configuration optimization model by adopting an improved genetic-greedy strategy transport capacity optimization algorithm of domain knowledge guided coevolution to obtain an optimal transport capacity configuration strategy, including an optimal allocation scheme of each transport capacity source, a selection strategy of ship types, a contract signing number and a route allocation plan. According to the method, the coupling model of the task requirements and the transport capacity resources is constructed, collaborative allocation of multi-source transport capacity is realized under the condition that the transport price is uncertain, the adaptability and the utilization efficiency of a transport capacity structure are improved, decision support is provided for enterprises, and reasonable allocation of the transport capacity resources in a dynamic environment is realized.
Owner:TRANSPORT PLANNING & RES INST MINIST OF TRANSPORT +1

Lightweight attitude recognition model construction method and device and attitude recognition method

The invention discloses a lightweight attitude recognition model construction method and device and an attitude recognition method. The construction method comprises the following steps: defining an attitude recognition model; defining a deep search space; the deep search space defines a cascading number space of deep stacking modules in the posture recognition model; defining a fusion selection space according to the deep search space; the fusion selection space defines a selection strategy space for selecting a fusion object by a fusion layer in the posture recognition model; and in the deep search space and the fusion selection space, searching a lightweight attitude recognition model for attitude recognition. According to the method, the lightweight attitude recognition model with both precision and operation speed can be quickly searched, so that the lightweight attitude recognition model can be deployed on a small platform such as an unmanned aerial vehicle.
Owner:XIAN TECH UNIV

Federal learning optimization method based on adaptive loss threshold

PendingCN121525895AMachine learningData setFactor selection
According to the federated learning optimization method based on the self-adaptive loss threshold, the threshold is dynamically calculated and adjusted according to the error change trend in the training process, so that the client selection strategy is optimized. Secondly, the invention provides a two-factor selection strategy which can select different selection strategies according to different training stages. And finally, a historical information interaction mechanism is provided, so that the client selection is not only based on the current performance index, but also can be flexibly adjusted according to the change trend of the historical performance. According to the mechanism, the client selection process can adapt to the requirements of different training stages, and the training stability and convergence efficiency can be remarkably improved. Experimental results show that compared with a Pow-d method, the global model accuracy rates of the AHIBCS on the three public data sets of COVID-19, FMNIST and CIFAR-10 are improved by 2.1%, 2.4% and 2.7% respectively.
Owner:STATE GRID HENAN INFORMATION & TELECOMM CO

A method and system for selecting participants in crowd intelligence perception based on prediction of cooperation willingness

The present invention relates to the field of crowd intelligence sensing technology, and in particular to a crowd intelligence sensing participant selection method and system based on cooperation willingness prediction. It includes obtaining a participant data set, obtaining an estimated cooperation cost through platform tasks; constructing a multi-armed bandit problem model, virtually employing the participant data set, and calculating the participant pair exploration rate; setting a threshold for the participant pair exploration rate, processing the data set using a link prediction model to obtain a predicted cooperation cost, and obtaining a converged cooperation cost based on the employed worker pair data and the estimated cooperation cost or predicted cooperation cost; obtaining a participant set using a minimum cost priority participant selection strategy; and updating the estimated cooperation cost and the employed worker pair data. It solves the problem of how the platform selects participants to minimize the cost of task completion when the relationship between the participants is not completely known or there is no historical data of a large number of participant pairs cooperating.
Owner:YANTAI UNIV

Three-generation genome SV detection method based on deep learning

The invention discloses a three-generation genome SV detection method based on deep learning, and belongs to the technical field of structural variation detection. The problem that an existing structure variation detection method is poor in accuracy is solved. The method comprises the following steps: firstly, extracting variation features from a comparison result with a reference genome, obtaining a submatrix of a fragment according to the extracted variation features, then carrying out deep coding on the submatrix of the fragment through a convolutional neural network to obtain coding features, splicing the coding features of a plurality of fragments into a 2000bp feature matrix, and finally, extracting the feature matrix; a global dependency relationship among the fragments is captured through a Transform network, so that a variation region can be accurately identified in a longer sequence; after variation is detected, variation sites are subjected to clustering analysis, breakpoint positions are accurately positioned, an automatic support reading selection strategy is designed, support readings can be automatically screened according to comparison quality, false positive areas are further filtered, and the reliability of a detection result is ensured. The method disclosed by the invention can be applied to SV detection in the third-generation genome.
Owner:HARBIN INST OF TECH +1