Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

42 results about "Global iteration" patented technology

Self-adaptive asynchronous federated learning method with local privacy protection

The invention discloses a self-adaptive asynchronous federal learning method with local privacy protection, which comprises the following steps that: a central server initializes a global model and broadcasts global model parameters, a gradient cutting standard, a noise mechanism and a noise variance to all participating users; each user firstly uses samples extracted from local data to train a global model and cuts and disturbs the gradients one by one, then the disturbed gradients are sent to a central server, the central server selects the first K disturbed gradients from a buffer queue to perform average aggregation, the averaged gradient is substituted into a stochastic gradient descent formula to update global model parameters, and a gradient cutting standard, a noise variance and a learning rate are adaptively adjusted according to the number of iterations in a preset stage; and then the central server broadcasts the updated global model parameters, the gradient cutting standard and the noise variance to K users participating in updating in the last round, and the local users and the central server repeat the operation until the number of global iterations reaches a given standard.
Owner:XI AN JIAOTONG UNIV

Coupling calculation method of deepwater gas-liquid two-phase flow circulating temperature and pressure

The invention discloses a coupling calculation method of deepwater gas-liquid two-phase flow circulating temperature and pressure. The coupling calculation method comprises the following steps: (1) calculating the vertical coordinate of a grid; (2) applying the initial condition; and (3) from the beginning of an original value, carrying out iterative calculation on node temperature and pressure data of drill string drilling fluid and annulus drilling fluid in sequence, until the temperature and the pressure reach the condition of convergence, finishing the iteration, and saving and outputting the last-time iterative calculation result as the final deep sea gas-liquid two-phase flow shaft temperature and pressure simulation result, wherein the iteration is named as a global iteration. The coupling calculation method improves the calculation precision.
Owner:CHINA UNIV OF PETROLEUM (EAST CHINA)

Electric-energy optimal control method for intelligent micro-grid with energy storage device

ActiveCN104022503ALow costReduce peak-to-valley load differencesForecastingAc network load balancingSmart microgridDynamic planning
The invention discloses an electric-energy optimal control method for an intelligent micro-grid with an energy storage device. The method comprises the steps that relevant parameters are initialized; global iteration is started, evaluation network weights and executive network weights are initialized; local iteration is started, an evaluation network and an executive network are trained by utilizing a self-adapting dynamic planning method, and neural network weights are corrected, wherein the evaluation network is used for approximating to an optimal performance index function, and the performance of the current battery control strategy is evaluated by utilizing the evaluation network weights; the executive network is used for approximating to an optimal control strategy and minimizing total cost in the primary global iteration; whether the current local iteration is completed or not is judged, if not, the local iteration is repeated, and if yes, an iteration performance index function and an iteration control law are updated to obtain the optimal solution; whether the current global iteration meets convergence precision or not is judged, if not, the global iteration is repeated, if yes, an optimal battery control strategy is obtained according to the optimal performance index function, and power utilization cost is calculated.
Owner:INST OF AUTOMATION CHINESE ACAD OF SCI

Turbo-Equalization Methods For Iterative Decoders

Certain embodiments of the present invention are improved turbo-equalization methods for decoding encoded codewords. In one embodiment, in global decoding iteration i, the magnitude values of all decoder-input LLR values (Lch) are adjusted based on the number b of unsatisfied check nodes in the decoded codeword produced by global iteration i−1. The improved turbo-equalization methods can be used as the sole turbo-equalization method for a given global decoding session, or interleaved with other turbo-equalization methods.
Owner:AVAGO TECH INT SALES PTE LTD

Identified association graph self-optimization mechanism

The invention provides an identified association graph self-optimization mechanism based on deep learning, and the mechanism combines a knowledge graph technology and a deep learning technology, carries out the identification processing of continuously collected data through a deep learning model, and then adds the data to an association graph. Data processing is carried out through a plurality ofdistributed data storage nodes and distributed computing nodes in the association graph, and a series of global iterations are carried out through three parts of local computing, communication unitsand fence synchronization on the basis of a block synchronization parallel computing model. Self-adaptive dynamic optimal allocation of computing resources is realized according to the resource utilization rate, the processing performance and the locality of the data of the system. And carrying out continuous disambiguation analysis and clustering calculation on the data added into the associationgraph to carry out continuous simplification and correction so as to realize continuous self-optimization of the association graph.
Owner:CHINA UNIV OF PETROLEUM (EAST CHINA)

Large-scale graph data processing method

The invention discloses a large-scale graph data processing method used for solving the technical problem that an existing large-scale graph data processing method is low in efficiency. According to the technical scheme, each parallel task is converted into a series of global iterative tasks, and iteration at every time is divided into three ordered stages, namely computation, global communication and barrier synchronization. The computation is divided into global communication and local communication, and the local computation includes a series of continuous internal iterations. The global communication stage includes that each working node sends a message of a current global iteration to a working node of a next global iteration. The barrier synchronization includes that a master node waits for that message passing of all the working nodes is completed, and then starts the next global iteration. Since multiple internal iterations are adopted during processing, times of the global iterations are decreased. Iteration times needed for highway data processing single-source shortest paths in the northeast of United States are decreased from more than 3800 times in the background art to 25 times; execution time is shortened from 1200s in the background art to 60s.
Owner:NORTHWESTERN POLYTECHNICAL UNIV

Elevator system self-learning optimal control method and system based on deep reinforcement learning

ActiveCN111753468AOptimal Elevator Control StrategyDesign optimisation/simulationConstraint-based CADElevator systemEngineering
The invention relates to an elevator system self-learning optimal control method and system based on deep reinforcement learning. The control method comprises the following steps: establishing an operation model and a probability distribution model; preprocessing the data information of the elevator system to obtain current data information; carrying out global iteration is according to the current data information, and carrying out local processing through iteration of a plurality of asynchronous threads: for each asynchronous thread, training a local action evaluation network through deep reinforcement learning according to the current data information, and correcting the weight of the action evaluation network; determining a global action evaluation network according to the weight of the action evaluation network until multi-thread iteration and global iteration are finished; and obtaining an optimal elevator control strategy according to the global action evaluation network so as to determine the average waiting time. In the global iteration process, local processing is carried out through iteration of the plurality of asynchronous threads, the weight of the action evaluation network is determined, and the optimal elevator control strategy is obtained through self-learning.
Owner:INST OF AUTOMATION CHINESE ACAD OF SCI

Apparatus and method for breaking trapping sets

An error correction data processing apparatus includes a noise predictive calibration circuit operable to calibrate a first set of filter coefficients based on a first data set and a second set of filter coefficients based on a second data set, and includes a first noise predictive detector operable to receive the first set of filter coefficients. The apparatus further includes a decoder operable to perform a first global iteration with the first noise predictive detector and determine a violation check count value, and a second noise predictive detector operable to receive the second set of filter coefficients if the violation check count value is less than a predetermined value or receive the first set of filter coefficients if the violation check count value is greater than the predetermined value.
Owner:BROADCOM INT PTE LTD

Remote sensing image segmentation method adopting region splitting technology

ActiveCN104657995AReduce the effect of speckle noiseImprove computing efficiencyImage enhancementImage analysisMarkov fieldContext model
The invention discloses a remote sensing image segmentation method adopting a region splitting technology. The method comprises the following steps: firstly, an image is subjected to initial segmentation with a watershed segmentation algorithm, and image regionalization is realized; a region adjacency graph is established for the regionalized image; the region adjacency graph is modeled based on the Markov field and is subjected to k-means initial marking; a iteration part is started, the initially marked image is subjected to Gibbs sampling marker optimization and initial region merging based on a global iteration weight of a Markov model, and record of a merging process is protected in a binary tree mode simultaneously; then the initial segmentation image is subjected to region splitting, and returning to initial regionalization configuration is performed; according to the positive correlation between the node number of a binary tree structure of a region and dimension of an object in a scene, the dimension weight in each regional space context model is adjusted adaptively, region markers are updated, and a final segmentation result is obtained. The noise influence can be eliminated very well, and adaptive marker updating in complicated scene of the image can be realized.
Owner:HEFEI UNIV OF TECH

SOVA sharing during LDPC global iteration

A technique for decoding information, including a Viterbi decoder configured to decode (1) Front-end processed ADC data and (2) an output of an iterative error correction decoder in the event error correction decoding fails. The iterative error correction decoder is configured to decode Viterbi decoded data generated by the Viterbi decoder.
Owner:SK HYNIX MEMORY SOLUTIONS

Systems and Methods for Power Measurement in a Data Processing System

Various embodiments of the present invention provide systems and methods for data processing. As an example, a data processing circuit is disclosed that includes: a data detector circuit, a data decoder circuit, and a power usage control circuit. The data detector circuit is operable to apply a data detection algorithm to a data input to yield a detected output. The data decoder circuit is operable to apply a data decode algorithm to a data set derived from the detected output to yield a decoded output. The power usage control circuit is operable to force a defined number of global iterations applied to the data input by the data detector circuit and the data decoder circuit regardless of convergence of the data decode algorithm.
Owner:AVAGO TECH INT SALES PTE LTD

Lattice code detection method and system

InactiveCN110276226AMeet the needs of image processing with different resolutionsEliminate distractionsImage enhancementImage analysisDot matrixSlide window
The invention discloses a lattice code detection method and system, and the method comprises the steps: carrying out the preset binarization processing of a grayscale image of a target image, wherein the preset binarization processing represents the binarization processing of a global iteration threshold adopted according to the definition of the grayscale image, and / or the preset binarization processing represents the adaptive local binarization processing; performing filtering processing on the binary image; traversing the binarized image after filtering processing according to the determined sliding window, calculating the average energy density of a selected area of the sliding window, and determining a candidate area according to an energy density threshold value; calculating the average energy density of the corroded candidate regions, and determining a plurality of first regions according to an energy density threshold; calculating the energy density among the sub-frames of the first area and the overall energy density of the first area, and determining a target area; and carrying out merging processing on the target area to obtain a target dot matrix code area. Different dot matrix positioning requirements can be met, and the decoding accuracy is improved.
Owner:BEIJING VISION BRILLIANCE TECH CO LTD

Energy consumption control method for wireless sensor network

The invention provides an energy consumption control method for a wireless sensor network, through which the energy consumption of the wireless sensor network can be reduced. The method comprises thefollowing steps: carrying out global iteration, that is, initializing network weight evaluation; carrying out local iteration, that is, training an evaluation network self-adaptive dynamic planning, and carrying out selection according to a received system status vector so that the evaluation network can generate an output meeting the lowest performance indicator standards, wherein the evaluationnetwork is used for approximating the optimum system performance indication function, and a current control strategy is an initially determined optimum control strategy; judging whether the local iteration is completed or not; if the local iteration is completed, updating a performance indicator function according to a current output value of the evaluation network, updating the optimum control strategy according to the initially determined optimum control strategy, and judging whether the current global iteration is completed or not; and if the current global iteration is completed, setting the updated optimum control strategy as the optimum control strategy for the energy consumption of the wireless sensor network. The invention relates to the field of wireless communication.
Owner:UNIV OF SCI & TECH BEIJING

Grid deviation space-time adaptive processing method based on local grid splitting

The invention discloses a grid deviation space-time adaptive processing method based on local grid splitting. The method comprises the following steps: acquiring a space-time adaptive processing dictionary and a clutter covariance matrix of a clutter space; initializing an orthogonal projection matrix and a noise projection matrix of the clutter space, and initializing a space-time steering vector set, a clutter atom subscript set, a global iteration number, a global iteration threshold, a local iteration number, a local iteration error threshold and a local maximum iteration number; searching space-time steering vectors of clutters in the space-time adaptive processing dictionary, forming a space-time steering vector set of an updated clutter space by the space-time steering vectors, and calculating a noise projection matrix of the updated clutter space; and according to the updated noise projection matrix of the clutter space, obtaining a space-time adaptive processing weighting vector of grid mismatch. According to the method, the problem that the estimation precision is related to the number of local grid division is solved, the operand is reduced, and the method has better processing efficiency.
Owner:XIDIAN UNIV

Hypersonic wind tunnel diffuser and design method thereof

The invention discloses a hypersonic wind tunnel diffuser and a design method thereof. The diffuser comprises an original diffuser body and further comprises a detachable built-in diffuser body installed in the original diffuser body. The design method of the diffuser comprises the steps: selecting the axial installation distance, the contraction angle, the diameter of an equal straight section, the length-diameter ratio and the divergence angle of the built-in diffuser as design parameters based on the size of an original diffuser and the size of a spray pipe outlet; obtaining a three-dimensional configuration based on the composition relation of the original diffuser and the built-in diffuser; generating a grid based on the three-dimensional configuration, and selecting a typical wind tunnel state corresponding to the spray pipe with the minimum outlet diameter to perform anti-back-pressure efficiency simulation; adjusting design parameters of the built-in diffuser through an optimization algorithm, and carrying out the global iteration optimization by taking improvement of anti-back-pressure efficiency as a target. The diffuser is simple in structure and can flexibly adapt to test requirements of the hypersonic wind tunnel. According to the design method of the diffuser, the design sample space can be greatly enriched, and the defect that design is carried out purely depending on experience is overcome.
Owner:中国空气动力研究与发展中心超高速空气动力研究所

Sparse ant system supporting large-scale combination optimization problem and solving method

The invention provides a sparse ant system supporting a large-scale combination optimization problem. The system comprises an initial solution construction module, an iterative solving module, a sparse pheromone matrix update module and an optimal solution calculation module. The invention furthermore provides a method for calculating an optimal solution of the large-scale combination optimizationproblem by using the sparse ant system. The method comprises the steps of 1, constructing an initial solution; 2, calculating optimal solutions by multi-time iteration, calculating global iterative optimal solutions in M ants, circulating each ant for N times, obtaining a sub-step solution node j during each calculation, finally obtaining feasible solutions of current iteration, selecting the global optimal solution in the feasible solutions of the M ants, and performing comparison and update on the global optimal solution of the current iteration and the iterative optimal solution so far; 3,after each iteration is finished, updating a sparse pheromone matrix according to the global iterative optimal solution; and 4, performing loop iteration according to rules of the steps 2 and 3, until the defined iterative frequency or total cost of the feasible solutions is smaller than or equal to a target optimal solution.
Owner:BEIJING UNIV OF POSTS & TELECOMM

Lattice code positioning method and system

The invention discloses a lattice code positioning method and system. The lattice code positioning method comprises the steps: carrying out preset binarization processing on a grayscale image of a target image, wherein the preset binarization processing represents the binarization processing of a global iteration threshold adopted according to the definition of the grayscale image, and / or the self-adaptive local binarization processing; performing filtering processing on the binary image; traversing the binarized image after filtering processing according to the determined sliding window, calculating the average energy density of a selected area of the sliding window, and determining a candidate area according to an energy density threshold value; calculating the average energy density of the corroded candidate areas, and determining a plurality of first areas according to an energy density threshold; calculating the energy density among the sub-frames of the first area and the overall energy density of the first area, and determining a target area; and carrying out merging processing on the target area to obtain a target dot matrix code area. Different dot matrix positioning requirements can be met, and the decoding accuracy is improved.
Owner:BEIJING VISION BRILLIANCE TECH CO LTD

Parallel asynchronous particle swarm optimization method and system and electronic equipment

The invention provides a parallel asynchronous particle swarm optimization method and system and electronic equipment, and the method comprises the steps: building a fitness function for a to-be-optimized target, and enabling the fitness function to be used for measuring a decision variable; grouping the particle groups, and randomly initializing initial positions, optimal values and diversity optimization parameters of particles in each particle group; establishing an information sharing mechanism for each particle group, wherein the information sharing mechanism is used for sharing the optimal value of each particle group; arranging the particle groups on different CPU cores for distributed parallel iterative calculation, and asynchronously updating historical optimal values of the particle groups according to an information sharing mechanism; and when the global number of iterations is greater than or equal to the threshold value of the number of iterations, ending the iterative update of each particle group, and outputting the optimal value of the particle group as a final optimization result. Compared with a traditional particle swarm algorithm, the method has the advantages that the optimization performance and robustness of the algorithm are improved, the calculation amount of the algorithm is reduced, the operation efficiency of the algorithm is improved, and the method can be suitable for various complex optimization scenes.
Owner:SHANGHAI JIAO TONG UNIV

Method for optimizing large-scale cloud service process

The invention relates to a method for optimizing large-scale cloud service process. The optimization method comprises the following steps: randomly distributing a plurality of execution plans to a plurality of parallel nodes; enabling each parallel node to execute local iteration in parallel to process an execution plan in the parallel node until the local iteration is stopped, enabling the parallel node to need to process the execution plan by using a discrete empire butterfly optimization algorithm and a dependency-conflict repair algorithm during each local iteration; gathering the execution plans obtained after the local iteration processing in all the parallel nodes is finished together; judging whether a global iteration stopping condition is met or not, and if yes, directly outputting an optimal execution plan; otherwise, redistributing the aggregated execution plan to a plurality of parallel nodes, and then repeating the steps. The solving efficiency and the solving precision of the cloud service process optimization method are improved, and the method can be effectively suitable for large-scale cloud service process optimization problems with different service process structures, QoS constraints and service association constraints.
Owner:SUZHOU UNIV

Systems and Methods for Power Monitoring in a Variable Data Processing System

Various embodiments of the present invention provide systems and methods for data processing. As an example, a data processing circuit is discussed that includes: a data detector circuit, a data decoder circuit, and a power monitor circuit. The data detector circuit is operable to apply a data detection algorithm to a data input and a decoded output to yield a detected output. The data decoder circuit is operable to apply a data decoding algorithm to the detected output to yield the decoded output. The power monitor circuit is operable to receive a first power status signal from the data detector circuit and a second power status from the data decoder circuit, and to calculate a power usage of a combination of at least the data detector circuit and the data decoder circuit. In such a system, a number of global iterations through a combination of the data decoder circuit and the data detector circuit is variable and both of the first power status signal and the second power status signal varies at least in part as a function of the number of global iterations.
Owner:LSI CORPORATION

Asynchronous federated learning method and device for improving utilization efficiency of edge device, and medium

The invention provides an asynchronous federated learning method and device for improving the utilization efficiency of edge equipment and a medium, which can reduce the data competition of a plurality of threads in a server on a global model and improve the concurrency performance of the server, and comprises the following steps: the edge equipment meeting a model data transmission condition actively requests the global model from the server; if the state of the event object is false, the server sends the global model to the edge device through the dispatcher component; the edge devices meeting the conditions train a global model through local data to obtain a local model; a collector component of the server enqueues the local model into a queue; the updater component pops up the local model and the shadow model from the queue to execute aggregation operation, assigns an aggregation result to the shadow model, assigns the value of the shadow model to the global model when a dequeue count value reaches a set value, and updates the number of global iterations; and continuously iterating until a set global total iteration number is reached.
Owner:NAT UNIV OF DEFENSE TECH

Elevator system self-learning optimal control method and system based on deep reinforcement learning

ActiveCN111753468BOptimal Elevator Control StrategyDesign optimisation/simulationConstraint-based CADElevator systemEngineering
The present invention relates to a self-learning optimal control method and system of an elevator system based on deep reinforcement learning. The control method includes: establishing an operation model and a probability distribution model; preprocessing the data information of the elevator system to obtain current data information; Perform global iteration according to the current data information, and perform local processing through multiple asynchronous thread iterations: for each asynchronous thread, according to the current data information, use deep reinforcement learning to train the local action evaluation network, and correct the weight of the action evaluation network; until multiple At the end of the thread iteration and the end of the global iteration, the global action evaluation network is determined according to the weight of the action evaluation network; the optimal elevator control strategy is obtained according to the global action evaluation network to determine the average waiting time. In the global iterative process, the present invention performs local processing through a plurality of asynchronous thread iterations, determines the weight value of the action evaluation network, and obtains the optimal elevator control strategy through self-learning.
Owner:INST OF AUTOMATION CHINESE ACAD OF SCI

Space-time adaptive processing method and device based on dictionary correction and storage medium

The invention discloses a space-time adaptive processing method and device based on dictionary correction and a storage medium. The method comprises the following steps: initializing a space-time guide dictionary; estimating the power of atoms in the space-time steering dictionary through an iterative adaptive algorithm, and obtaining space-time vector atoms most related to clutter points; acquiring a corresponding true clutter Doppler frequency based on the determined space-time vector atoms; replacing the Doppler frequency of the determined space-time vector atoms with the true clutter Doppler frequency; and performing global iteration through an iterative adaptive algorithm, atoms matched with clutter ridges are obtained, and a new space-time oriented dictionary [theta](k+1) is formed. According to the method, the clutter ridge broadening problem caused by off-grid can be effectively relieved, and the clutter suppression performance of sparse recovery STAP is improved; and the method does not depend on exact environment priori knowledge, the algorithm performance is not sensitive to parameter selection, the CNCM can be accurately estimated only through few training samples, and the calculation complexity and the calculation amount are greatly reduced.
Owner:INNER MONGOLIA UNIV OF TECH

A method for optimizing the layout of roadside communication units in a dedicated short-range communication system for Internet of Vehicles

The invention discloses a layout optimization method of a short-range communication system roadside communication unit special for the Internet of vehicles. The method includes the steps that firstly, an Internet-of-vehicles roadside unit network in an actual application scene is extracted and abstracted into a complex network; secondly, a target function and an efficiency and cost function are built; thirdly, initial disturbance is introduced to simulate the cascading failure process of the network, and parameters of the target function are adjusted; fourthly, parameters in the CRO algorithm are determined, and the iteration process is executed to obtain the global optimal solution; fifthly, change trends of four theoretical index values corresponding to the global iteration process are recorded and saved in an information database, and qualitative and quantitative evaluation and reference are provided for similar network optimization. The network structure of the Internet-of-vehicles roadside unit network is optimized by means of related theories of the complex network according to the common cascading failure problem in a communication network, and reliable guarantee is provided for normal operation of a vehicle self-organizing network.
Owner:BEIHANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products