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

607 results about "Iteration process" patented technology

Iteration is the repetition of a process in order to generate a (possibly unbounded) sequence of outcomes. The sequence will approach some end point or end value. Each repetition of the process is a single iteration, and the outcome of each iteration is then the starting point of the next iteration.

Intelligent agent system optimization method and device based on intelligent fault analysis and cross-generation knowledge inheritance

The invention relates to an intelligent agent system optimization method and device based on intelligent fault analysis and cross-generation knowledge inheritance, and belongs to the technical field of artificial intelligence. According to the method, interaction abnormal signals are captured in real time by deploying a lightweight log probe, and a tool benefit prediction model based on reinforcement learning is constructed to automatically generate an improvement proposal when the failure rate exceeds a threshold value; an agent genealogy map is established to realize automatic inheritance of a new agent on core memory and abandonment of failure knowledge, and a disastrous forgetting blocker is deployed to dynamically extract a functional module from a genealogy to deal with key capability degradation. Aiming at the problems of fault response lag, knowledge inheritance fracture, key capability degradation and the like in an intelligent agent system iteration process, the invention creatively provides a cooperation mechanism of an intelligent fault analysis layer and a cross-generation knowledge inheritance network, and the fault self-healing capability, version stability and service continuity guarantee level of the system are remarkably improved.
Owner:KUNLUN YUAN ARTIFICIAL INTELLIGENCE TECHNOLOGY (SHANGHAI) CO LTD

Text generation method, electronic device, storage medium and computer program product

The invention discloses a text generation method, electronic equipment, a storage medium and a computer program product, and relates to the field of large model technology and natural language processing. The method can comprise the steps that in response to an input instruction acting on an operation interface, first text data corresponding to the input instruction are determined, and the first text data comprise a plurality of first lexical elements; on the basis of the first text data, second text data are generated in a multi-iteration mode, in the same iteration process, multiple second lexical elements in the second text data are generated on the basis of the first text data and candidate lexical elements in the first text data, and the candidate lexical elements are used for representing lexical elements corresponding to the same iteration process; and displaying the second text data on the operation interface. The technical problem of low text generation accuracy is solved.
Owner:ALIBABA CLOUD FEITIAN (HANGZHOU) CLOUD COMPUTING TECH CO LTD

Parallel iteration Runge-Kutta method for simulating second-order nonlinear differential equation and application

The invention discloses a parallel iteration Runge-Kutta method for simulating a second-order nonlinear differential equation. The parallel iteration Runge-Kutta method comprises the steps that a simulated ordinary differential equation is converted into a target nonlinear implicit differential equation, and a system differential algebraic equation is established; judging whether a system matrix and a decomposition iteration matrix in the system differential algebraic equation are updated or not; a four-order extrapolation method is adopted to obtain a predicted differential variable derivative, and if a prediction result is not converged, parallel iterative operation is carried out through a four-stage four-order Runge-Kutta method; after the iteration process is converged, judging whether an operation result accords with a preset error range, if not, updating a system matrix according to a step size reduction strategy and re-decomposing an iteration matrix, and if yes, updating the operation result; and storing and displaying an operation result when judging that simulation is finished. The algorithm stability and the calculation efficiency of the rigidity problem can be considered.
Owner:SHANGHAI JIUTONGFANG TECHNOLOGY CO LTD

Non-correlation parallel machine scheduling method based on deep reinforcement learning

The invention belongs to the technical field of industrial intelligence, and relates to a deep reinforcement learning-based non-correlation parallel machine scheduling method, which comprises the following steps of: constructing a mathematical model suitable for non-correlation parallel machine scheduling; the machining time of each workpiece on each heterogeneous machine is collected, and normalization processing is carried out; initializing a genetic algorithm scheduling population and a depth Q network; constructing a deep reinforcement learning training framework; expressing a state vector by using the average fitness, the optimal fitness and the optimal individual code; operating parameters of the genetic algorithm are controlled by using the action space, and parameters of the deep Q network are updated by using a reward function; and dynamically controlling operator selection in a genetic algorithm iteration process by using the trained deep reinforcement learning model to obtain an optimal scheduling solution and realize scheduling of the non-correlation parallel machine. According to the method, a deep reinforcement learning algorithm is provided for solving similar problems in the manufacturing industry production scheduling field by analyzing a non-correlation parallel machine scheduling problem data model, and the production efficiency is improved.
Owner:DALIAN UNIV OF TECH

Three-dimensional point cloud target recognition attack-resisting method based on self-adaption imperceptibility

The invention relates to an adaptive imperceptible three-dimensional point cloud-based target identification confrontation attack method. The method comprises the following steps of: obtaining a central point from an original point cloud based on the magnitude of a saliency and imperceptibility joint score; gaussian perturbation is carried out on the original point cloud based on the obtained center point to generate a confrontation point cloud, and function mapping is carried out on a dynamic interval of Gaussian perturbation in the process so as to avoid sudden change of perturbation and improve the attack success rate of the confrontation point cloud; constructing a loss function for iterative training, the loss function constraining disturbance through a joint regularization item, and adaptively and dynamically adjusting a regularization item weight parameter in an iteration process; continuously iteratively generating confrontation point clouds, and optimizing the confrontation point clouds by taking minimization of confrontation loss under constraint as a target so as to improve the imperceptibility of disturbance; and finally generating a three-dimensional confrontation point cloud. According to the method, good balance between the point cloud attack resistance imperceptibility and the resistance strength can be realized.
Owner:JIMEI UNIV +1

Verification method and device based on genetic algorithm

The embodiment of the invention provides a verification method and device based on a genetic algorithm, and the method comprises the steps: randomly selecting a plurality of groups of initial key parameters from a preset key parameter value range, and calculating the fitness of each group of initial key parameters in a processor verification process, processing the plurality of groups of initial key parameters according to the fitness, a preset selection probability, a preset crossover probability and a preset mutation probability to finally obtain a plurality of groups of target key parameters, taking the plurality of groups of target key parameters as the plurality of groups of initial key parameters, executing the steps again until a preset termination condition is met, and stopping iteration to obtain a plurality of target key parameters; and calculating and outputting a processor verification coverage rate according to the plurality of groups of target key parameters and the key parameter value range. The target key parameters are obtained based on the genetic algorithm, local optimum is avoided, the parameters are continuously optimized in the iteration process, the high processor verification coverage rate can be obtained, and the processor verification efficiency and quality are improved.
Owner:SHANDONG YUNHAI GUOCHUANG CLOUD COMPUTING EQUIP IND INNOVATION CENT CO LTD

Pipeline residual strength prediction method and system based on artificial intelligence algorithm

The invention discloses a pipeline residual strength prediction method and system based on an artificial intelligence algorithm, and the method comprises the steps: constructing a CNN-BiLSTM-Adaboost model; the CNN-BiLSTM-Adaboost model comprises a CNN feature extraction module, a BiLSTM time sequence module and an Adaboost optimization module, wherein the CNN feature extraction module is used for extracting features of the corroded pipeline according to parameter data of the corroded pipeline, the BiLSTM time sequence module is used for obtaining a feature time sequence of the corroded pipeline by utilizing a bidirectional LSTM propagation mechanism, and the Adaboost optimization module is used for adjusting sample weights in an iteration process to generate a strong predictor; and during prediction, the residual strength of the pipeline is predicted through the trained CNN-BiLSTM-Adaboost model. The method is high in prediction precision and wide in adaptability.
Owner:NANJING FORESTRY UNIV

Expensive constrained multi-objective optimization method and system based on constrained hyper-volume expectation

The invention discloses an expensive constrained multi-objective optimization method and system based on constrained hyper-volume expectation. The method comprises the following steps: firstly, initializing a population and training an agent model; then calculating a correlation coefficient between the solution with the best population convergence and the constraint violation degree of the solution, and performing classification processing; then, optimization is carried out on the agent model, and a new solution is selected by utilizing a filling criterion based on constraint hyper-volume expectation to carry out evaluation and is filled into a file; further selecting and screening individuals by using an improved environment to enter a next-generation population; and finally, judging whether a stop condition is reached or not, if so, outputting a final population, otherwise, turning to execute an iteration process. According to the method, the optimization efficiency is remarkably improved, local optimum is avoided, constraint conditions are precisely processed, and the population can be more effectively converged to a better feasible region.
Owner:CHINA UNIV OF GEOSCIENCES (WUHAN)

Constraint optimization method for solving heterogeneous task allocation of multiple unmanned aerial vehicles

The invention belongs to the technical field of intelligent optimization of intelligent robots, and particularly relates to a constraint optimization method for solving heterogeneous task allocation of multiple unmanned aerial vehicles. According to the method, a composite real number coding mode is adopted, and more information can be recorded in short individual codes; a self-adaptive alpha constraint level comparison strategy is adopted in the algorithm, relative feasible solution information can be effectively utilized in the iteration process, the convergence speed of the alpha level is adaptively influenced through the proportion of a relative feasible solution, the safety level is gradually improved, the search range of an optimal allocation scheme is limited within a satisfactory scheme, and the optimal allocation efficiency is improved. Therefore, an effective distribution scheme can be planned.
Owner:BEIJING UNIV OF TECH

High-performance low-complexity LDPC (Low Density Parity Check) decoding method for satellite communication

The invention discloses a high-performance low-complexity LDPC (Low Density Parity Check) decoding method for satellite communication, and belongs to the field of digital signal processing. And performing iterative decoding on the received channel information based on a low density parity check (LDPC) code in the 5G standard. The implementation method comprises the following steps of: dynamically setting an erasing range by utilizing the amplitude of current check node information when a variable node is updated; the amplitude of check node information is gradually increased along with the increase of the number of iterations, the whole iteration process is divided into two parts, and the erasing range is set by adopting different parameters. And outputting the decoding result when the decoding result is correct or reaches the maximum number of iterations. According to the method, the reliability of variable node information is improved through the two points, reliable variable information is prevented from being erased, unreliable variable node information is prevented from being spread, and excellent decoding performance is achieved. Non-linear operation in a log-likelihood ratio belief propagation algorithm LLR-BPA does not exist in the decoding algorithm, and the decoding algorithm has the advantages of being high in decoding performance and low in decoding complexity.
Owner:BEIJING INST OF TECH

Distributed training method, device, system, equipment and medium

The invention provides a distributed training method, device and system, equipment and a medium, and relates to the technical field of artificial intelligence, in particular to a distributed machine learning technology. The method comprises the following steps: calculating a current compression ratio and a current delay step length according to a bandwidth parameter and a transmission delay parameter of a current network; compressing the gradient of each round of calculation in the iteration process according to the current compression ratio to obtain a compression gradient, and updating an accumulative error term; when the number of iterations reaches the current delay step length, calculating a current update gradient according to a current compression gradient and an accumulative error term; and uploading the current update gradient to a server, so that the server performs aggregation according to the current update gradient uploaded by each computing node to obtain an aggregation gradient, and calculates global parameters of a model according to the aggregation gradient and issues the global parameters to each computing node. According to the method, the model precision is not lost in the process of reducing the communication overhead, so that the training efficiency and training precision of the model are improved.
Owner:BEIJING FACE WALL INTELLIGENT TECHNOLOGY CO LTD

Multiplying active code decoding method based on SO-ORBGRAND decoder

The invention belongs to the technical field of communication, and particularly relates to a multiplication active code decoding method based on an SO-ORBGRAND decoder. The SO-ORBGRAND decoder outputs soft information by calculating a correct posterior probability of a conjecture code word sequence, and high-performance decoding is realized through soft information exchange and optimization processing between the row decoder and the column decoder. In the iteration process, a method for judging the validity of the code word through the check matrix is designed by utilizing the polarization characteristic of the multiplication active code, so that the iteration is supported to jump out in advance to reduce the complexity. The invention provides a multiplication active code decoding algorithm based on an SO-ORBGRAND decoder, and compared with various non-system multiplication active code decoding algorithms and the performance of traditional long-code polarization codes, the method has remarkable advantages in error correction performance; a method for judging the validity of the code word through the check matrix by utilizing the polarization characteristic design of the multiplication positive code is provided, the complexity is reduced while the judgment success rate is improved, and the implementation overhead is reduced.
Owner:UNIV OF ELECTRONICS SCI & TECH OF CHINA

Method for optimizing logistics sorting encoder error based on improved grey wolf algorithm

The invention belongs to the technical field of electric digital data processing, and particularly relates to a method for optimizing errors of a logistics sorting encoder based on an improved grey wolf algorithm. According to the method, Logistic chaotic mapping and Gaussian perturbation are superposed to generate a diversity initial parameter population so as to break through the limitation of traditional random initialization, a fitness function is designed to quantify an angle compensation residual error, and parallel computing is utilized to accelerate evaluation. In the iteration process, global exploration and local development are dynamically balanced through adaptive convergence factors, a bimodal perturbation mechanism is constructed in combination with a differential evolution strategy and Levy flight variation, population effectiveness is maintained through reflection boundary processing, guiding of # imgabs0 # wolf is enhanced through dynamic weight distribution, the position of a leader wolf is updated by adopting an elitist retention strategy, and the population effectiveness is improved. And finally, outputting the optimal compensation parameter when the maximum number of iterations or the residual threshold is met, thereby improving the positioning precision and the operation efficiency of the logistics sorting system.
Owner:QILU UNIVERSITY OF TECHNOLOGY (SHANDONG ACADEMY OF SCIENCES) +2

Multi-objective optimization method and system for aviation airborne product

The invention relates to the technical field of intelligent manufacturing, and discloses a multi-objective optimization method and system for an aviation airborne product, so as to improve the overall performance of the aviation airborne product. The method comprises the steps of obtaining an initial population based on a series of initial model files of a target aviation airborne product, and determining a constraint condition and a target function of the target aviation airborne product; the constraint condition comprises the range of each iteration parameter; in each iteration process of obtaining the global optimal Pareto frontier according to the constraint condition, performing non-dominated sorting and congestion degree calculation on the tth generation of population through an algorithm, selecting individuals with excellent performance from a sorting result, and applying gradient-guided Gaussian disturbance to generate a (t + 1) th generation of new solution; the tth generation of population comprises cross progenies of the (t-1) th generation of population; and when the target iteration termination condition is met, iteration is terminated, and the global optimal Pareto front is obtained.
Owner:HUNAN VANGUARD SCI & TECH CO LTD +1

Toxic text collection method and system based on retrieval enhancement generation

The invention relates to a toxic text collection method and system based on retrieval enhancement generation, and the method comprises the steps: firstly obtaining target text data through a crawling platform, manually constructing a small-scale initial data set through a cold start mode, and injecting the initial data set into a knowledge base as basic data; and performing semantic retrieval on each batch of target texts, obtaining the first k texts with semantic similarity from the knowledge base, reasoning by adopting a plurality of large language models through thinking chain reasoning in combination with the texts, generating a toxic label, and labeling the target texts to be labeled in the current batch. And injecting the labeled target text into a knowledge base, continuously collecting and labeling data in an iteration mode, and continuously optimizing the performance of a retriever through an incremental learning mechanism in the iteration process so as to realize toxic text collection. According to the method, large-scale, fine-grained and high-consistency toxic text tagging corpora can be efficiently accumulated, and the problems of high tagging cost, inconsistent quality and poor expansibility in traditional toxic text data collection are remarkably relieved.
Owner:JIANGXI UNIVERSITY OF FINANCE AND ECONOMICS

Dynamic S box generation method and test system based on spatiotemporal chaotic system

The invention discloses a dynamic S-box generation method and test system based on a spatiotemporal chaotic system, and the method comprises the steps: constructing an iteration matrix through employing a preset m sequence, and determining a coupling object index and a disturbance symbol of each grid in an iteration process; wherein each element of the iteration matrix represents a grid; performing dynamic disturbance on the initial chaotic sequence according to the coupling object index, the disturbance symbol, the index offset and the grid coordinate to obtain a dynamic chaotic sequence; initializing the two-dimensional coupling image lattice by using a dynamic chaos sequence, and carrying out iteration to obtain an iterated coupling image lattice; removing the previous N iteration results of the coupled image lattice after iteration; and performing numerical conversion on the chaotic sequence value obtained by the (N + 1) th iteration to construct an S box. According to the S box design and test method, the safety and the attack resistance of the S box are improved, and an efficient and reliable research tool can be provided for cryptography researchers.
Owner:BEIJING ELECTRONICS SCI & TECH INST

Power load prediction method based on NIWPSO + CNN + LSTM + Attention

The invention relates to the field of computers, in particular to an NIWPSO + CNN + LSTM + Attention-based power load prediction method, which comprises the following steps of: acquiring power data, and performing preprocessing and standardization operation on the power data; the preprocessed data are used as input, and a CNN-LSTM-Attention model is constructed; carrying out hyper-parameter optimization on the CNN-LSTM-Attention model by utilizing an NIWPSO (Non-International World Weighted Particle Swarm Optimization) algorithm; setting a residual error monitoring and dynamic updating mechanism, monitoring the residual error in real time during model prediction, and automatically triggering a retraining process once the residual error exceeds a preset threshold value; a nonlinear variation inertia weight strategy is innovatively adopted, dynamic adjustment is carried out according to an algorithm iteration process and a particle search state, a PSO algorithm is prevented from being caught in a locally optimal solution too early, meanwhile, a mutation thought of a genetic algorithm is used for reference, mutation operation is carried out on particles in a one-dimensional mode, the algorithm is assisted to efficiently obtain a globally optimal solution in a complex solution space, and model parameter optimization is guaranteed.
Owner:NANJING UNIV OF POSTS & TELECOMM

Method and device for simplifying cue word set and storage medium

The invention relates to the technical field of data processing, and discloses a method and device for simplifying a cue word set and a storage medium, the method is applied to a server, the method comprises the steps that a preset initial cue word set is input into a large language model for learning, and in each iteration process, the cue word set is obtained; according to the first score value and / or the second score value of the current round, whether the number of the initial input data is changed is determined, and each round of iteration process comprises the following steps that a preset first algorithm is adopted to score each piece of first synthetic data, and the first score value is obtained; and / or, a preset second algorithm is adopted to score each piece of second synthetic data to obtain a second score value, and the second synthetic data is obtained by learning each piece of initial input data included in the initial cue word set by applying the large language model added with the scene constraint condition. Therefore, the prompt word better conforms to the current application scene, and the quality of the generated synthetic data is improved.
Owner:CHINA TELECOM CORP LTD TECHNOLOGY INNOVATION CENTER +1

Equipment deployment method and device, computer equipment and storage medium

The invention discloses an equipment deployment method and device, computer equipment and a storage medium, and the method comprises the steps: building an equipment deployment optimization target, and determining a target function and a constraint condition; based on the equipment deployment information of the deployed project, key parameters are predicted through a preset neural network, and the key parameters are used as initial parameter values of a quantum ant colony algorithm; iterating the quantum ant colony algorithm for a preset number of iterations, and based on an objective function, calculating an objective function value corresponding to each equipment deployment scheme meeting the constraint condition to obtain a local optimal equipment deployment scheme generated in each iteration process; and selecting a global optimal equipment deployment scheme from all the local optimal equipment deployment schemes. The optimization target is solved through the neural network and the quantum ant colony algorithm, the equipment deployment scheme is determined, the labor cost can be reduced, and the equipment deployment efficiency can be improved.
Owner:SHENZHEN JIESHUN SCI & TECH IND

Intelligent questioning and answering method and system for knowledge graph in electric power marketing field based on large model Agent

The invention discloses an intelligent questioning and answering method and system for a knowledge graph in the field of power marketing based on a large model Agent, and the method comprises the steps: obtaining a to-be-queried question and the large model Agent, enabling a question rewriting tool of the large model Agent to complete and rewrite the question Q based on a constructed prompt word module and a pre-trained templated processing model, and carrying out the quality verification of a rewritten new question Q '; inputting the new question Q'passing the quality auditing into a large model Agent, and converting the new question Q'into at least one query statement q; the query statement q is processed through a result generation tool, a result R returned by processing is received, and the result R contains data requested by the query statement q; the large model Agent evaluates the result R, if the result R is correct, the result R is output, if not, the iteration process is started, and in the iteration process, the large model Agent judges whether to rewrite the generated new question Q'and generate a new query statement.By means of the question query rewriting method and device, the question query rewriting capacity can be enhanced, the query result accuracy can be improved, and the query efficiency can be optimized.
Owner:STATE GRID JIANGSU ELECTRIC POWER CO LTD MARKETING SERVICE CENT +1

Method for quantizing neural network model, data processing method, and related apparatuses

Provided in the embodiments of the present specification are a method for quantizing a neural network model, a data processing method, and related apparatuses. A regularization item is introduced into an original loss function, a minimized loss function is taken as an objective function, an optimization expectation is taken as a constraint condition and, during a plurality of iteration processes, some first weights among a plurality of first weights are sparsified into second weights, so as to achieve the aim of sparsifying some first weights into the second weights while ensuring that the sum of the loss function of a target neural network model and the regularization item is as small as possible, such that the target neural network model can take both operation efficiency and calculation precision into account; that is, on the basis of expanding the scenarios to which the target neural network model can be applied, good precision of the target neural network model is ensured.
Owner:HANGZHOU ALICLOUD FEITIAN INFORMATION TECH CO LTD

System and method for separating a quantum state into multiple subspaces

A quantum computer system comprises first and second registers which are used to separate a quantum state into multiple subspaces of a 2n dimensional Hilbert space. A quantum state comprising 2n elements is defined on the first register, the first register comprising n qubits; a quantum state is defined on the second register, the second register comprising one or more qubits; and receiving a value k, where k is a binary integer such that 0 =< k =< n. The method further performing a bit-wise iteration process comprising: (i) performing a quantum entanglement between the registers to separate the quantum state on the first register into distinct subspaces of the Hilbert space which are indexed by the entangled values on the second register, and (ii) measuring an outcome on the second register to find a match with a portion of k, wherein said portion of k increases incrementally with the iteration process until the match is with all of k. The bit-wise iteration is used to separate the quantum state on the first register into subspaces of the Hilbert space, wherein elements of the subspaces have different Hamming weights, and one subspace contains only elements of Hamming weight k.
Owner:QUANTINUUM LTD

Active power distribution network energy storage optimization configuration method and system based on improved parrot algorithm

The invention discloses an active power distribution network energy storage optimization configuration method and system based on an improved parrot algorithm, and the method comprises the steps: building a multi-target energy storage optimization configuration model with node voltage deviation, energy storage cost and power distribution network loss as target functions according to a preset constraint condition; a chaos theory is introduced to initialize a parrot algorithm population, in each iteration process, weights of foraging behaviors, staying behaviors, communication behaviors and stranger afraid behaviors in the parrot algorithm are dynamically adjusted according to fitness values of individuals and current iteration times, Gaussian variation or Cauchy variation is selected according to variation probabilities, and a parrot algorithm is obtained. Mutation operation is carried out on the parrot individuals in the iteration process, and finally an improved parrot algorithm is obtained; and solving the multi-target energy storage optimal configuration model according to the improved parrot algorithm, and outputting an energy storage optimal configuration result. Efficient collaborative optimization of multiple objective functions is realized, and more accurate and comprehensive decision support is provided for energy storage optimization configuration of the active power distribution network.
Owner:NANCHANG UNIV

Repeated construction project cost optimization method

The invention discloses a repetitive construction project cost optimization method, and particularly relates to the field of project construction. Comprising the following steps: S1, parameter setting: initializing the starting time of a time window and the number of algorithm iterations; s2, generating an initial construction plan; s3, if the starting time of the time window is larger than the difference between the construction period of the initial construction plan and the length of the time window, the starting time of the time window is set to be 0; s4, generating a corresponding relaxation problem by adopting a rolling time domain method; s5, solving the relaxation problem in the step S4 by using a constraint programming method; s6, updating the number of iterations, and updating the starting time of the time window; and S7, repeating the steps S3-S6 until an iteration condition is met, and outputting an optimal solution in an iteration process. By the adoption of the technical scheme, the problem that an existing construction plan compilation method with cost as an optimization target cannot be suitable for different construction scenes is solved, and meanwhile, when medium and large-scale cases are solved, an approximate optimal solution can be obtained within reasonable time.
Owner:BEIJING JIAOTONG UNIV

Power distribution network reactive power scheduling scheme generation method and device, computer equipment, readable storage medium and program product

The invention relates to a power distribution network reactive power scheduling scheme generation method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of randomly generating a plurality of power distribution network reactive power scheduling schemes based on parameters of a power distribution network system, and establishing a fitness function; based on a fitness function, iteratively optimizing the plurality of schemes, each iteration process comprising: based on fitness values of the plurality of schemes of the last iteration, determining a target scheme of the current iteration; based on the target scheme and a piecewise nonlinear decreasing weight strategy, performing first optimization, second optimization and horizontal crossover operation on the plurality of schemes of the last iteration to obtain a plurality of schemes of the current iteration, and based on the fitness function, determining fitness values of the plurality of schemes of the current iteration; and when the number of iterations reaches a preset number of iterations, taking the optimal scheme of the last iteration as an optimized power distribution network reactive power scheduling scheme. The method can prevent from falling into a local optimal solution during optimization.
Owner:MAOMING POWER SUPPLY BUREAU GUANGDONG POWER GRID CORP

Multi-strategy segmented fusion adaptive SPGD algorithm for laser coherent combination system

The invention belongs to the field of coherent combination, and discloses a multi-strategy segmented fusion adaptive SPGD algorithm for a laser coherent combination system. The method is used for solving the problems that in a traditional SPGD algorithm, due to the fact that a fixed gain coefficient and disturbance amplitude exist, the convergence speed and stability of the algorithm are contradictory, and the optimal performance is difficult to achieve in different environments and system states. According to the method, the algorithm is divided into three stages in the iteration process, and self-adaptive adjustment of the gain coefficient and the disturbance amplitude is realized according to the characteristics of a hardware system in combination with three different iteration strategies. According to the method, the convergence speed of the algorithm can be remarkably improved, meanwhile, the stability of the algorithm can be effectively improved, the algorithm has higher robustness in consideration of hardware system characteristics, and the good application prospect and feasibility of the algorithm are shown.
Owner:GUANGDONG UNIV OF TECH

Micro-grid scheduling method

The invention provides a microgrid scheduling method, and the method comprises the steps: constructing a first optimization model which comprises a plurality of objective functions and a plurality of constraint conditions; converting the first optimization model into a second optimization model in a QUBO form; the second optimization model is solved through a mixed quantum approximate optimization algorithm to obtain a micro-grid scheduling strategy, the mixed quantum approximate optimization algorithm comprises a quantum approximate optimization algorithm and a classical algorithm, the quantum approximate optimization algorithm is used for solving the second optimization model through quantum evolution in the multi-round iteration process, and the classical algorithm is used for solving the second optimization model through quantum evolution in the multi-round iteration process. The classical algorithm is used for adjusting optimization parameters of the quantum approximate optimization algorithm based on candidate solutions generated by quantum evolution in each iteration process, parallel exploration of multiple potential solutions is achieved, and therefore the probability of finding an optimal solution is improved within feasible time.
Owner:STATE GRID XINJIANG ELECTRIC POWER CO ECONOMIC TECH RES INST

Federal learning method, system and device based on prototype and storage medium

The invention discloses a federal learning method, system and device based on a prototype and a storage medium, in the iteration process of the last round, a client transmits model parameters to train a target model, the prototype of the client is transmitted to a server, the server carries out model aggregation and prototype aggregation, and the target model is obtained. Finally, prediction is carried out based on the distance between a local prototype and a global model, on a heterogeneous data set, a prototype which is aggregated as a feature space can effectively integrate feature expressions from data of different distributions, the negative influence of data heterogeneity is reduced, the model reasoning performance is improved under the condition of data heterogeneity of different clients, and the model reasoning efficiency is improved. The technical problems that due to the fact that client data of financial institutions often have large differences in the aspects of storage devices, dispersion degrees and the like, data distribution is not independent and identical due to the differences, the basic assumption of a standard machine learning algorithm cannot be met, the model reasoning performance is poor, and the federation model prediction accuracy is low are solved.
Owner:PING AN TECH (SHENZHEN) CO LTD

Heterogeneous cluster task scheduling fusion method based on Q learning and genetic algorithm

The invention discloses a heterogeneous cluster task scheduling fusion method based on Q learning and a genetic algorithm, and belongs to the technical field of intelligent scheduling. Collaborative optimization is carried out through a dynamic feedback mechanism of Q-Learning and the global search capability of the genetic algorithm: in an initialization stage, a cluster state space and an action space are defined, and a multi-Q-value table and an initial population are generated; during task scheduling, nodes are selected based on a # imgabs0 #-greedy strategy, reward values are calculated, and a Q value table is updated; meanwhile, the scheduling scheme is coded into chromosomes, a new population is generated through roulette selection, crossover and variation, and a Q value table is optimized; in the iteration process, the convergence is judged according to the Q value change or the population fitness change, and the strategy is dynamically adjusted. According to the method, the advantages of double algorithms are fused, local optimum is avoided, the task processing efficiency and the resource utilization rate are remarkably improved, adaptive strategy updating during cluster state change is supported, and the method is suitable for an efficient task scheduling scene of a large-scale heterogeneous cluster.
Owner:ARTIFICIAL INTELLIGENCE INNOVATION RES INST OF ZHEJIANG UNIV OF TECH BINJIANG DISTRICT HANGZHOU

Cable-strut tension structure active control method and device based on physical information neural network

The invention relates to an active control method and device for a cable-strut tension structure based on a physical information neural network. The method comprises the following steps: acquiring the randomly generated actuation amount of a plurality of units; calculating a tangent stiffness matrix of the structure after the prestress and the actuation amount are applied, and when the matrix is a positive definite matrix, fitting structural displacement by using a physical information neural network; calculating the unit internal force of the corresponding structure based on the maximum displacement value obtained through fitting, judging whether the obtained unit internal force and the maximum displacement value meet constraint conditions or not, if yes, calculating a loss function value, further judging whether the maximum number of iterations is reached or not, and if not, calculating a loss function value; guiding the iteration process of the physical information neural network by using a loss function value in the next iteration; and taking the action amount corresponding to the minimum displacement of the structure fitted in previous iterations as a final required result, and realizing active control based on the final required result. Compared with the prior art, the method has the advantages of high calculation efficiency, wide application range and the like.
Owner:SOUTHEAST UNIV