69 results about "Parallel algorithm" patented technology

The invention relates to the technical field of data processing, in particular to a new energy uncertainty multi-type energy storage staged capacity optimization configuration method, which comprises the following steps: acquiring energy storage parameters, new energy output and user load data through edge equipment, and verifying photovoltaic conversion efficiency and lithium battery electrochemical behaviors through a physical model to generate a high-confidence data set; the federal cooperative system fuses distributed node data to output a health degree distribution diagram, and the memory network analyzes the temperature and the charge-discharge coupling effect to generate a dynamic compensation coefficient alpha; an optimization engine receives the alpha value and wind and light prediction, generates a staged scheme through a parallel algorithm, and inputs the staged scheme into a simulation system for verification; and the edge node executes the scheme and collects response data, and twin compares a simulation value with an actual deviation to update a collection strategy. The method eliminates data distortion through physical rule verification, corrects aging unit state deviation through a dynamic compensation mechanism, optimizes decision binding real-time health degree, and effectively defends charging and discharging analysis errors of a virtual power plant dispatching desk.

The invention discloses a cloud edge end heterogeneous resource cooperative training strategy based on hierarchical hybrid parallel, and aims to research how a cross-domain distributed computing framework fully utilizes various cloud edge end heterogeneous computing resources, research an equipment graph and computing graph construction mechanism, construct a simplified and merged computing graph and equipment graph, and reduce the search space of the hierarchical hybrid parallel strategy. A search algorithm of the hierarchical hybrid parallel algorithm is researched, a possible hierarchical hybrid parallel algorithm is searched through a dynamic programming algorithm and a resource constraint mechanism, and efficient hierarchical hybrid parallel in a cloud edge-end collaborative cross-domain scene is realized; a cross-domain hierarchical hybrid parallel optimization strategy during operation is studied, the calculation speed between nodes is adjusted through calculation of data loads in clusters and secondary balanced distribution of the calculation loads, and slow nodes during operation are avoided. By means of the mechanism, efficient adaptation of the cloud edge heterogeneous computing resources and the hybrid parallel strategy is achieved, the utilization efficiency of the cloud edge heterogeneous computing resources is improved, and the cross-domain distributed training speed of the model is increased.

The invention discloses a large power grid security check calculation method suitable for multiple time scales, and belongs to the technical field of data processing, and the method comprises the following steps: S1, obtaining the calculation content of a multi-user security check task, evaluating the complexity of the calculation content, and selecting a parallel calculation mode based on an evaluation result; s2, based on a parallel computing device pool set in a parallel computing mode and a job scheduling strategy, extracting computing sub-tasks of the multi-user security check task, and dynamically distributing the computing sub-tasks to corresponding computing nodes by adopting a load balancing mechanism; and S3, processing the calculation subtasks of the plurality of calculation nodes by adopting a parallel algorithm of a CPU + GPU hybrid architecture, merging calculation subtask results of the plurality of users, and generating a final security check report. According to the scheme, the computing resource utilization rate of the resource pool is improved, and the large-scale power grid safety checking efficiency is remarkably improved.

The invention discloses an efficient and parallel PIC / MCC rapid calculation method and system. According to the method, a PIC / MCC calculation process is divided into a charge distribution module, an electric field solution module, a particle propulsion module and a collision processing module, and parallel execution is carried out on a GPU through an independent kernel function. All simulation data are resident in a GPU video memory, and storage access is optimized by adopting a structured array and row main sequence layout. GPU atomic operation is introduced in charge distribution to ensure data consistency; selecting a parallel algorithm to accelerate Poisson equation solution according to dimensions in electric field solution; an electric field is obtained through interpolation in particle propulsion, and the particle motion state is updated; and adopting a Monte Carlo method to judge the collision type in parallel in the collision processing. Synchronous control among the modules is realized through a GPU event mechanism, and self-consistency of a physical process is ensured. According to the method, the bottleneck of traditional serial calculation is broken through, the speed-up ratio of more than 30 times is realized while the calculation precision is kept, and the efficiency and expandability of plasma numerical simulation are remarkably improved.

Disclosed in the present invention is a runoff forecasting method based on an EEMD-ANN and meteorological factors and using a multi-core parallel algorithm. The method comprises: S1, performing feature screening, involving: on the basis of historical data of previous inflows and meteorology, performing screening of lag terms of the previous inflows by means of a partial autocorrelation function, using an inverse distance weighting method to reduce the weight of meteorological data, using a cross-correlation function and an index of agreement to perform screening of lag terms of rainfall, then performing feature selection by means of a correlation coefficient γ, and finally, using a correlation coefficient τ to perform screening of input factors; S2, performing data pre-processing, involving: using an empirical mode decomposition technique to decompose the input factors, performing data normalization processing, and dividing data into a training set and a test set; and S3, performing model forecasting and assessment, involving: integrating the input factors and inputting same into a forecast model, then performing parameter optimization by means of a parallel algorithm, outputting a forecast result, and performing model assessment. The present invention effectively improves forecast accuracy, accelerates the efficiency of model parameter calibration, and reduces model training time.

The invention provides a self-adaptive collaborative parallel optimization aviation complex structural member production scheduling method and system and a program product. The method comprises the following steps: presetting parallel algorithm related parameters; initializing a population, wherein the initialized population comprises a first scheduling scheme generated by using a random greedy heuristic algorithm; performing evolutionary optimization on the initialized population or the derivative population of the initialized population in parallel by utilizing a plurality of configured sub-threads to generate a new scheduling scheme; according to a comparison result of the current CPU utilization rate and the memory utilization rate and the target CPU utilization rate and the target memory utilization rate, adjusting the number of sub-threads which are running; according to the method, the problems that in the prior art, a scheduling scheme is not high in quality and low in optimization efficiency, and the utilization rate of computing resources is difficult to consider are effectively solved, the high-quality and high-adaptability scheduling scheme can be generated, the solving time is shortened, the computing resources are fully utilized, and the scheduling efficiency is improved. The production efficiency and the resource utilization rate are improved.

The invention discloses a data assimilation analysis system based on an object-oriented framework, and the system comprises the steps: obtaining assimilation parameters, defining an analysis variable space for the assimilation parameters through an unstructured grid framework module, and carrying out the distribution collection and boundary exchange calculation through a geometric grid and parallel algorithm communication module MPDD, a control variable, an observation variable, a covariance matrix, various transformation operators and a solver module are constructed, a multivariable price function and a price function gradient are established, the system transmits the multivariable price function and the price function gradient to a universal minimization solver, an analysis field is made to approach an observation result through iterative solution, and the price function gradient is obtained. Other points except the observation position are diffused outwards in the mode of the covariance matrix to influence other mode grid points, and an assimilation analysis field result is obtained. According to the method, rapid expansion and easy replacement of various different space variables, observation operators, covariance matrix schemes, multivariable solvers and the like are realized, and the flexibility and adaptability of a data assimilation analysis system are improved.

The invention discloses a transaction processing method and device in a block chain, electronic equipment and a storage medium, and belongs to the technical field of network security. In the method, after a request end obtains target transaction information required by an AMM transaction, transaction authenticity proof is carried out by using an anti-parallel algorithm; according to the method, a transaction request carrying part of target transaction information is sent to a consensus node in a block chain network, then supplementary transaction information of the transaction request is sent to the consensus node, and the supplementary transaction information comprises the remaining part of target transaction information, verification auxiliary information and authenticity proof information obtained by using an anti-parallel algorithm; a consensus node performs transaction authenticity verification based on verification auxiliary information and authenticity proof information, and responds to a transaction request based on an authenticity verification result, and when an anti-parallel algorithm is utilized to perform transaction authenticity proof, an attacker is difficult to forge a transaction matched with the authenticity proof, so that sandwich attacks can be effectively discriminated.

The invention relates to the technical field of environmental science, and discloses an analogue simulation method and system for large-scale gas diffusion. The method comprises the following steps: determining a simulation purpose and collecting related data, constructing a geometric model of a simulation area, carrying out multi-scale grid division, estimating a gas diffusion path, speed and concentration through preliminary simulation, then correcting an estimation result and selecting a proper physical model, and using a multi-scale coupling strategy to simulate the gas diffusion path, speed and concentration. The method comprises the following steps: carrying out coupling modeling on physical models of different scales, designing a parallel algorithm to simulate a multi-scale coupling model, evaluating a simulation result and analyzing an error in combination with actually measured data, carrying out optimization, analyzing the simulation result and evaluating an environmental risk, and displaying a gas diffusion process through three-dimensional visualization. According to the method, the simulation precision, the environmental adaptability and the response flexibility are improved, and precise simulation and risk assessment of large-scale gas diffusion in a complex environment are realized.

The invention discloses a power transmission line lightning stroke risk accelerated assessment system and method and a storage medium, and the method comprises the steps: obtaining the potential of each point in a space electric field of a to-be-assessed region according to a parallel algorithm, carrying out the simulation calculation of thundercloud and a downlink pilot model of the to-be-assessed region, and obtaining the ground thunderbolt distribution of the to-be-assessed region; according to the ground-to-ground lightning density of the to-be-evaluated region and the back-strike trip-out rate sub-model, the shielding failure trip-out rate sub-model and the inductive lightning trip-out rate sub-model, calculating the back-strike trip-out rate, the shielding failure trip-out rate and the inductive lightning trip-out rate of the power transmission line of the to-be-evaluated region respectively; and combining and calculating the back strike trip-out rate, the shielding failure trip-out rate and the inductive lightning trip-out rate of the power transmission line in the to-be-evaluated region to obtain an average value of the lightning trip-out rates of the base towers of the power transmission line in the to-be-evaluated region, and evaluating the lightning risk of the to-be-evaluated region through the average value of the lightning trip-out rates. By adopting the method, a lightning stroke risk assessment result can be obtained in a shorter time, and the lightning protection capability and the operation safety of the power system are improved.

The invention discloses a GPU (Graphics Processing Unit) parallel acceleration method for subdivision volume parameter evaluation, which comprises the following steps of: S1, storing and traversing a volume grid through a data structure containing a half side and a half surface, and judging the type of each volume unit; s2, performing CC body subdivision under any topology once, and isolating irregular vertexes if a plurality of irregular edges in a body unit must converge at an irregular point; s3, aiming at different topologies, numbering control grid vertexes by using a data structure containing a half surface and a half side, obtaining a subdivision matrix according to a numbering sequence and a CC subdivision rule, and pre-calculating a subdivision matrix feature structure; and S4, performing parallel parameter evaluation on different types of volume grid units. According to the method, the optimized parallel algorithm is designed, the multi-core architecture and the floating-point operation capability of the GPU are fully utilized, and the parameter evaluation speed of the complex body subdivision model is effectively improved.

The invention provides a distributed unstructured grid topology construction method, and relates to the field of unstructured grids, and the method comprises the steps: judging the dimension of an unstructured grid; if the grid is a three-dimensional grid, sequentially constructing 16 grid topologies consisting of points, edges, surfaces and a body on the basis of the body-point topology; if the grid is a two-dimensional grid, nine grid topologies composed of points, edges and surfaces are constructed in sequence on the basis of the surface-point topology; if the grid is a one-dimensional grid, sequentially constructing four grid topologies consisting of points and edges on the basis of the edge-point topology; when the grid topology is constructed, cross-process adjacency information is supplemented and synchronized through communication. According to the method, the problems of large memory demand, unbalanced load, repeated calculation and high communication cost during use in the prior art can be solved, a complete distributed unstructured grid topology construction parallel algorithm process is provided, and it is ensured that the method can be efficiently applied to different numerical simulation scenes.

The application discloses a kind of super-resolution SIM-FRET image reconstruction methods based on GPU parallel acceleration, including steps: FRET original image is collected using acquisition card and is stored;Program initialization is carried out;FRET single-channel original image is obtained and transmitted asynchronously to GPU global memory;Parallel image preprocessing is carried out to FRET single-channel original image group by GPU;Normalizing separation image is iterated by the heterogeneous computing of CPU and GPU Correlation Parameter Estimation, and the modulation intensity and initial phase of illumination light vector are estimated;GPU parallel SR-SIM linear wiener reconstruction;Super-resolution FRET three-channel image is obtained;FRET three-channel super-resolution image is processed based on channel sensitization intensity measurement method E-FRET data by GPU.The application improves the parallel algorithm structure suitable for GPU acceleration, and the heterogeneous computing of CPU and GPU is used for the iterative correlation parameter estimation algorithm which cannot be completely parallelized, improves the GPU histogram summation algorithm of biological image, and improves the time resolution of super-resolution SIM-FRET image reconstruction.

A layered hybrid parallel cloud edge end heterogeneous resource collaborative training strategy, research on how to fully utilize the multi-layered hybrid parallel strategy of the search space of the search algorithm of the layered hybrid parallel algorithm, through the dynamic programming algorithm and the resource constraint mechanism to search the possible layered hybrid parallel algorithm, realize the efficient layered hybrid parallel of the cloud edge end collaborative cross-domain scene; research on the runtime cross-domain layered hybrid parallel optimization strategy, through the secondary balanced allocation of the data load and the calculation load in the calculation cluster, adjust the calculation speed between each node, avoid the emergence of slow nodes at runtime. Through the above mechanism, the efficient adaptation of cloud edge heterogeneous computing resources and hybrid parallel strategy is realized, the utilization efficiency of cloud edge heterogeneous computing resources is improved, and the model cross-domain distributed training speed is improved.

The application discloses a regional scale process model simulation precision optimization method, system and device, relates to the field of model optimization, and comprises the following steps: inputting a crop parameter combination to be optimized into a crop-water productivity model, determining result variables yield and canopy coverage; performing cross-validation on the result variables yield and canopy coverage to determine evaluation indexes; and constructing a multi-objective index evaluation optimization model for optimization of key parameters of a regional scale PCMs; wherein the multi-objective index evaluation optimization model takes the sum of Nash efficiency coefficients NSE of the canopy coverage and the result variable yield in the training set and the test set as an objective function, and takes a root mean square error RMSE as a constraint condition; the evaluation indexes are input into the multi-objective index evaluation optimization model, a parallel algorithm group composed of multiple meta-heuristic optimization methods is adopted to solve the multi-objective index evaluation optimization model, and the key parameters of the regional scale PCMs are optimized to obtain an optimal crop parameter combination.

The invention discloses a medium and long term power system partition and local power grid balance capability evaluation and early warning method and system. The method comprises the following steps: based on real-time power system model data, superposing boundary data such as power generation and application of a medium-and-long-term power system and equipment maintenance / commissioning / retreating information to generate a medium-and-long-term power system model and section data meeting stable operation conditions; identifying power grid power supply subarea power grids and local power grids and internal power equipment information of the power grids in each time period; a multi-period parallel algorithm is adopted to calculate power generation and receiving balance, positive and negative standby and power supply margins of the subarea power grid and the local power grid; and then the medium-and-long-term power system balance condition and the standby and power supply margin condition are evaluated, and medium-and-long-term partitioned and local power grid balance monitoring and early warning functions are provided. Medium and long term power grid balance capability analysis data is provided, monitoring of future power grid balance capability operation characteristic change trend is powerfully supported, power grid long-period stability analysis capability is improved, and future risk early warning capability is improved.

The application belongs to the technical field of neural network simulation and analog, and particularly relates to a large-scale cortex pulse neural network asynchronous parallel simulation algorithm based on GPU. The application utilizes the advantages of multi-thread and texture memory of a computing graphics card, combines the general form of a biological brain receiving external stimulation and the general connection mode between neurons in the cortex, designs an asynchronous parallel algorithm framework of GPU and CPU, GPU is responsible for parallel evolution of neuron dynamics equations and block parallel calculation of isotropic connection in a local network, CPU is responsible for processing anisotropic long-range connection, and different neuron dynamics equations and plasticity learning rules can be compatible. Compared with the prior art, the application can effectively improve simulation speed, provides a tool for simulating a biological brain cortex in a single computing node, and is suitable for a single node multi-graphics card and a multi-node multi-graphics card distributed operation model.

The invention discloses an optimization method for a sparse dense matrix multiplication parallel algorithm based on CSR characteristics, including CSR sparse matrix data buffer design, load balancing of buffer parallel data preloading, load balancing of dense matrix calculation tasks on threads, and temporary storage design of partial cumulative sums of calculation results. The beneficial effect of the invention is that by analyzing some memory access problems existing in the current SpMM parallel algorithm, buffer design is used as an entry point, and load balancing is used as a means to achieve the purpose of improving computing performance and improving computer resource utilization.

The application discloses an intersatellite link satellite data transmission task scheduling method and system and a storage medium, and comprises the following steps: acquiring a satellite set, a task target set and a ground station set; determining a network connection plan and a data transmission plan in each time slot; acquiring a task constraint condition, establishing an intersatellite link satellite data transmission task scheduling model, and taking the maximum complete task data value transmitted to the ground station after T time slots as the scheduling model; constructing a parallel algorithm group, acquiring an optimal algorithm and corresponding parameter configuration, and matching the optimal algorithm with a scene instance. The method constructs an integer programming model of the intersatellite link satellite data transmission task model, constructs a parallel algorithm group, searches for an optimal algorithm configuration in a machine learning mode, and overcomes the contradiction between the cost and quality of huge computing resources required by parallelism.

The present invention proposes a method, device and readable storage medium for processing and refining point cloud data. In view of the fact that the existing technology generates point clouds with poor accuracy, insufficient refinement and many noise points, the present invention proposes an innovative solution. It uses a unique noise removal step to accurately remove noise points by using the expansion grid and custom feature growth judgment; through the point cloud segmentation and recognition step, accurate segmentation is achieved based on calculations such as local normal vectors; linear fitting, segmentation merging and resampling steps are used to optimize the point cloud with a parallel algorithm; and then the point cloud is reorganized according to the scene through the point cloud reorganization step. This algorithm process has significant advantages in the fields of robotics and surveying and mapping, which can greatly improve the quality of point clouds, enhance the actual application effect, and provide an efficient, general and easy-to-implement strategy for data processing in related industries, which will effectively promote the development of point cloud technology and the deepening of practice.

This invention relates to the field of traffic safety technology, specifically to an efficient simulation method for quantum systems based on parallel order reduction. The method includes the following steps: receiving the physical parameters of the quantum system and simulation requirements, whereby the physical parameters include electron mass, potential well size, and initial wave packet parameters; and the simulation requirements include simulation physical time and accuracy requirements. Based on the physical parameters, a time-dependent Schrödinger equation describing the dynamic behavior of the quantum system is established, and the Schrödinger equation is rearranged into matrix form using the Kronecker product. This invention utilizes Arnoldi to reduce the order of the matrix-form Schrödinger equation. By reducing the order, the large matrix in the original space is projected into a relatively small subspace, improving the efficiency of the simulation solution. The reduced-order Schrödinger equation is solved using a time-parallel algorithm. This approach overcomes the time step limitation imposed by the CFL condition, ensuring that a stable solution can be obtained with fewer time steps.

A method and high-performance apparatus for creating an approximate maximum k-cut of a graph. In various embodiments, nodes and weighted edges in a graph are used to compute a partitioning of a graph such that the edge weight between partitions is maximized. In various embodiments, the method and apparatus use a greedy random construction of cuts for its first approximation, whose solutions are then subject to local search for better solutions, and local path-relinking for the final level of refinement. In various embodiments, one or more of the greedy random construction, local search, and path relinking are parallelized. A novel parallel algorithm is disclosed for perturbing the state of a cut to see if a better one is cheaply available.

The invention relates to the technical field of electromagnetic field simulation modeling, in particular to a magnetotelluric efficient parallel simulation method and device based on random path integration, and the method comprises the steps: rewriting a deterministic solution based on a solution matrix equation into a form based on random path integration; in the offline preparation stage, the number of times that random walk particles pass through each lattice point is recorded, a path matrix is generated, the number of times that all random walk particles pass through all lattice points of a Robin boundary is recorded, a boundary path matrix is generated, and preset model grids are interpolated to different random walk grids to generate an interpolation matrix; and in the online stage, obtaining a final calculation result according to the path matrix, the boundary path matrix and the interpolation matrix. Therefore, the problems of unit communication, load balancing and I / O bottleneck in parallel simulation of a large-scale electromagnetic positive problem caused by parallelization transformation of a non-parallel algorithm adopted by an existing deterministic simulation method based on equation set solution in the prior art are solved.

The present application relates to an electromagnetic anisotropic large-scale parallel forward modeling method based on a PDE solving base. The method includes: constructing an anisotropic electromagnetic model according to the constitutive relationship between the electric and magnetic fields when facing anisotropic media; the constitutive parameters of the constitutive relationship include the dielectric constant, magnetic permeability, and conductivity of the anisotropic media; inputting the configuration file information of the electromagnetic problem to be solved into the PDE solving base, and the PDE solving base loads the electromagnetic model plug-in according to the configuration file information, constructs an electromagnetic field solving module, divides the calculation area into multiple sub-domains, allocates processes from the process set to each sub-domain, and in the processes corresponding to each sub-domain, parallelly solves the anisotropic electromagnetic model of the sub-domain according to the multi-level multi-granularity parallel algorithm and the configuration information in the electromagnetic field solving module, and outputs the electromagnetic field results of the sub-domain; obtaining the forward modeling result according to the electromagnetic field results of each sub-domain. Using this method can improve the efficiency and reliability of large-scale electromagnetic forward modeling.

The invention discloses a physical model parallel computing method of a Rydberg atomic electric field intensity measurement system, belongs to the technical field of quantum measurement, and solves the problem of how to improve the computing efficiency of solving the physical model of the Rydberg atomic electric field intensity measurement system. According to the method, a physical model of the Rydberg atomic electric field intensity measurement system is established by using a Bloch equation; diagonalizing the time discrete matrix, and calculating a physical model of the Rydberg atomic electric field intensity measurement system by using a parallel algorithm; frequency sweeping is carried out on the probe light, a relational expression between the splitting size measured by the transmission spectrum and the RH frequency is established, the splitting size measured by the transmission spectrum is corrected, and the intensity of the electric field to be measured is obtained by combining an analysis and calculation formula of the transmission spectrum; according to the method, on the premise of solving the physical quantity evolution process in the Bloch equation, the calculation efficiency is improved, the physical time concept can be embodied, the convergence is higher, and one-time solving can be carried out.

The application discloses an intelligent optimization method based on a multiphase flow LBM parallel algorithm, and comprises the following steps: 1) constructing a data dependency relationship model of the multiphase flow LBM parallel algorithm; 2) constructing a memory access performance analysis model of the multiphase flow LBM parallel algorithm; 3) according to the data dependency relationship model, the characteristics of GPU memory access and the data read-write relationship, proposing a kernel fusion rule of the multiphase flow LBM algorithm and a performance change estimation method after fusion; 4) combining the fusion rule and the performance change estimation method into a grouping genetic algorithm to search for an optimal fusion scheme; and 5) optimizing according to the optimal fusion scheme and comparing the performance before and after optimization. The method comprehensively considers the data transmission and memory access characteristics of the CUDA kernel, explores the performance gain under various CUDA kernel fusion strategies by using an intelligent search algorithm, and finally obtains the optimal CUDA kernel fusion strategy, and the acceleration effect of the method is significantly improved compared with a traditional multiphase flow LBM parallel method.

The invention relates to a radioactive substance diffusion prediction method, in particular to a radioactive substance diffusion prediction method based on a random depletion and particle size parallel algorithm, and mainly solves the problems that in the prior art, space-time specific processing of a source item is rough, depletion is not considered in dry-wet settlement correction, the particle size is single, statistics is insufficient, and calculation is time-consuming. The method comprises the following steps: 1, setting source item parameters, meteorological parameters and parallel computing parameters; 2, preprocessing the parameters respectively; 3, performing diffusion calculation on the particles of each particle size in the radioactive substance at each moment by adopting a parallel algorithm to obtain position information of all the particles, and storing the position information; 4, performing statistics to obtain space grid time integral activity concentration values and ground grid settlement activity values of the radioactive substances corresponding to each moment, and storing the space grid time integral activity concentration values and the ground grid settlement activity values into a hard disk according to a moment sequence; and 5, calculating to obtain the total time integral activity concentration and the total sedimentation activity of the radioactive substance, thereby realizing diffusion prediction of the radioactive substance.

The application discloses a method for recognizing and judging the health condition of a dairy cow based on video skeleton, and relates to the technical field of video behavior recognition, and comprises the following steps: step one: a camera collects images and sends the images to a terminal for processing, manual labeling and detection of key points of a dairy cow skeleton; step two: a Bottom-up posture estimation method algorithm is used to obtain the posture of the dairy cow, the Bottom-up posture estimation method comprising feature extraction, landmark positioning, part grouping and tracking; feature extraction: a parallel algorithm of bilinear interpolation is used to process the images, BatchNormalization is used for batch data standardization, self-attention gating is used to highlight the significant features of the region of interest, and the U-net feature extraction network model is improved; the application applies the skeleton calibration technology to the dairy cow to recognize the behavior of the dairy cow, and the recognition accuracy is increased and the innovation is improved; compared with the skeleton data, other modalities will produce more calculation consumption, and the robustness is insufficient when facing complex backgrounds and human scale changes, view angle changes and motion speed changes.

The application discloses a kind of based on cloud computing's multi-source heterogeneous big data fast cleaning and fusion method, to solve the technical problems of poor adaptability of heterogeneous data in the prior art, low processing efficiency, insufficient fusion accuracy, flow batch processing fragmentation and lack of quality closed loop.The method is based on cloud native distributed architecture, steps include: building global semantic ontology model, realizing the unified access and standardization adaptation of multi-source heterogeneous data, combining consistent hash fragmentation and elastic computing power scheduling to complete data efficient distribution;Adaptive missing value repair, robust outlier detection and distributed bloom filter deduplication are performed using parallel algorithms, data cleaning is realized quickly;Through two-stage entity alignment and data source credibility weighted conflict resolution, multi-source data deep fusion is completed;After full-amount quality check, hierarchical storage is carried out, rule adaptive optimization is realized relying on reinforcement learning, and data consistency and low latency are guaranteed by supporting flow batch incremental processing mechanism.

A method and high-performance apparatus for creating an approximate maximum k-cut of a graph. In various embodiments, nodes and weighted edges in a graph are used to compute a partitioning of a graph such that the edge weight between partitions is maximized. In various embodiments, the method and apparatus use a greedy random construction of cuts for its first approximation, whose solutions are then subject to local search for better solutions, and local path-relinking for the final level of refinement. In various embodiments, one or more of the greedy random construction, local search, and path relinking are parallelized. A novel parallel algorithm is disclosed for perturbing the state of a cut to see if a better one is cheaply available.