38 results about "Recursive computation" patented technology

The invention discloses a cache optimization method for a cell-free multiple-input-output environment based on a VPPO algorithm, and relates to the technical field of edge cache optimization. Multiple independent environment instances are generated and run in parallel; training the strategy network based on the plurality of independent environment instances and the value network to obtain a trained strategy network; and inputting the cache proportion of the current edge server into the trained strategy network to obtain the adjustment of the cache proportion of the edge server. A complex dynamic scene is simulated through multi-environment parallel interaction, multi-scene data are collected in parallel, and the adaptability and decision-making ability of the model to dynamic changes are improved. According to the method, a non-Markov environment is adapted by using Retract advantage estimation, and the problem of deviation of traditional advantage estimation in such scenes is solved by truncation weight and recursive calculation and processing of the condition that state transition depends on history, so that the advantage function calculation is more accurate, a foundation is laid for strategy optimization, and the performance of a model in a complex dependency relationship scene is improved.

This invention discloses a two-end reliability assessment method and system for SDN based on OBDD. Leveraging the centralized control and easy programming characteristics of SDN, this method establishes an SDN two-end reliability assessment model. By designing the topology management module and path calculation module within the network model, a two-end minimum path set that meets bandwidth and latency constraints is obtained. A minimum path set-breadth-first search algorithm is used to determine the variable ordering of the BDD. The corresponding OBDD is constructed using the Shannon formula in the OBDD construction module. Finally, the reliability assessment module recursively calculates network reliability from the bottom up, completing the SDN two-end reliability assessment. This method combines methods such as the constrained adjacency endpoint matrix algorithm, the minimum path set-breadth-first search algorithm, and the ordered binary decision diagram (OBDD) to assess connectivity reliability in SDN, demonstrating good scalability.

The invention relates to a monitoring data recursive calculation method and application thereof in a CCUS project. An estimation error used for describing the difference between an estimation value and a true value and a measurement error used for describing the difference between a measurement value and the true value are introduced; the comprehensive error is obtained by dividing the previous estimation error by the sum of the previous estimation error and the current measurement error; an initial estimation value and an initial estimation error are randomly given, then an initial comprehensive error is estimated, then a subsequent estimation value is calculated through a recursive algorithm, and the initial comprehensive error is updated to be used for calculation of the subsequent estimation value. After a plurality of rounds of measurement and calculation, an estimated value is closer to a real value, and all measurement data in the earlier stage does not need to be backtracked, so that the problems of a large amount of repeated redundancy and large error of monitoring data are effectively reduced; and the method provided by the invention is used for realizing carbon dioxide leakage monitoring in a CCUS project.

The invention provides a binary program vulnerability mining method, device and system based on symbolic execution and taint analysis and a medium, and the method comprises the steps: disassembling a target program, extracting a control flow graph (CFG) and a data flow graph (DFG), recording a node state through a hash table, and generating a feature representation through nonlinear transformation; performing simulation execution based on CFG, DFG and feature representation, recording variable symbol values to obtain path conditions, and recursively solving constraints to generate path mapping; marking input as taint data, recursively calculating a propagation path to generate a taint flow diagram, and determining a taint state after sensitive operation; checking whether the stains are subjected to sensitive operation or not, and if the influence of integral formula calculation exceeds a threshold value, judging that potential vulnerabilities generate a candidate set; and calculating grades through a risk assessment formula, and generating a report containing positions, types, grades and repair suggestions. The method combines symbolic execution and taint analysis, can comprehensively and accurately detect vulnerabilities, has remarkable precision and efficiency advantages, and is suitable for complex program security analysis.

The invention relates to a pseudorandom sequence generation method and device and a medium, and the method comprises the steps: calculating weight vectors of different powers through modular operation based on a primitive polynomial of a Galois field, and obtaining a bit weight value of each bit; splitting the offset value into a plurality of accumulations in the form of power of 2 by using a binary decomposition method, and obtaining an initial vector at the offset position by combining the bit weight value at each bit; according to the bit width generation requirement, the one-dimensional weight matrix is expanded into a two-dimensional generation matrix based on the principle of a digital root method; based on the two-dimensional generation matrix, recursively calculating matrixes under different bit weights by using matrix power operation, and generating weight matrixes under different processing bit widths; generating a pseudorandom sequence starting point according to a mode of multiplying the initial vector at the offset position by the weight matrixes under different processing bit widths; and generating the pseudorandom sequence from the starting point of the pseudorandom sequence by using a recursive formula. According to the invention, waste of storage resources and bandwidth resources can be avoided.

The application discloses a kind of based on artificial intelligence's geotechnical engineering survey safety monitoring system and method, it is related to rock-soil mass deformation safety monitoring technical field, specifically including the following steps: obtaining multi-source data, including the deformation of rock-soil mass, stress and environmental data, model input data source is established;Data fusion processing is carried out;With the result of data fusion processing as input, through data simulation, the stress and deformation correlation of rock-soil mass is obtained;Based on deep learning, with the stress and deformation correlation of rock-soil mass as basement, through the method of time series prediction, short-term recursive calculation model and long-term trend analysis model are respectively constructed;Real-time multi-source data value is obtained, the displacement prediction vector of rock-soil mass is obtained according to the short-term recursive calculation model, compared with the preset threshold value, when exceeding the threshold value, real-time early warning is carried out, and the long-term trend analysis result is visualized.The application enhances engineering safety, improves the long-term stability and reliability of infrastructure.

This invention discloses a method, system, equipment, and medium for calculating pressure loss in perfluorohexanone (PFH) pipeline networks, relating to the field of computer-aided design technology. The method traverses the component tree of the pipeline network building information model using a three-dimensional geometry engine to form a set of pipeline network features; constructs a directed graph topology representing the flow direction of the pipeline network, calculates potential energy loss, converts local resistance into equivalent length, and constructs a weighted directional topology weight matrix; obtains hydraulic balance under single-phase flow conditions through iterative calculation of node simultaneous equations, and generates the corresponding static pressure distribution results of the pipeline network; recursively calculates the dryness and density changes of the gas-liquid mixture using the equation of state, introduces a two-phase flow friction multiplication factor for nonlinear compensation, outputs a three-dimensional dynamic compensation pressure field model, and performs rendering and digital judgment; this solution solves the problem of low accuracy in pressure loss calculation caused by the inability of existing technologies to accurately simulate the flash evaporation effect of PPHH, thus improving the reliability of pipeline network design.

The invention discloses a storage pool creation method and device, equipment, a medium and a product, which are applied to the technical field of storage, and comprise the following steps: obtaining the number of fault domain units and the number of copies; when the number of the fault domain units is greater than the number of the copies, judging whether each fault domain unit meets a first preset capacity balance condition or not based on the minimum effective capacity and the expected effective capacity corresponding to each fault domain unit; if the first preset capacity balance condition is not met, grouping recursive calculation is carried out on each fault domain unit to obtain a target effective capacity, and whether a second preset capacity balance condition is met is judged based on the target effective capacity and the expected effective capacity; and under the condition that a second preset capacity balance condition is not met, creating a main pool based on the target effective capacity and creating an auxiliary pool based on the residual capacity. Therefore, the utilization rate of storage resources can be improved.

The invention provides a time sequence coupling constraint embedded power resource scheduling method and system and a storage medium, and relates to the technical field of energy management, and the method comprises the steps: obtaining a classical security constraint economic scheduling model; recursive calculation is carried out on the classic security constraint economic dispatching model; wherein in the forward propagation process of the recursive calculation process of the neural network model, parameters of the neural network at all time points are kept consistent, and meanwhile, the output of the calculation unit of the current time step serves as the input of the calculation unit corresponding to the next time step so as to connect all the calculation units; embedding the constraint condition into an output end of a set neural network model, and constructing a gradient return information highway; and training a set neural network model by using the power training data, and adjusting neural network parameters to obtain a power scheduling model so as to output a power resource scheduling scheme. Through the coupling calculation unit, the parameter sharing and the recursive structure, the calculation efficiency is remarkably improved, and the power resource scheduling scheme can be quickly generated.

The application discloses a kind of man-machine mapping and deft operation method and system based on virtual hand drive, form complete closed loop from hand perception to deft hand control.It includes: image is collected with monocular RGB camera, after normalization etc., 21 three-dimensional coordinates of joint of hand are predicted through multi-task deep neural network, then data is smoothed through dynamic filtering;Particle swarm optimization algorithm is used to fit the shape parameters of MANO hand model;Based on weighted singular value decomposition, the rotation matrix of root joint is calculated, and the global rotation matrix of non-fingertip joint is calculated recursively combined with hand topology structure;Local coordinate system of hand is established, and angle mapping function is designed to convert joint angle into deft hand control quantity;Through double buffering mechanism interpolation processing low frame rate perception result, hybrid control is realized by fusing current data.The scheme improves precision, robustness, adaptability and safety, and provides technical support for efficient and reliable human-computer interaction.

This application relates to a nonlinear recursive calculation and monitoring method for relative blood volume. The method includes acquiring prior baseline parameters, discrete laboratory data sequences, and real-time physiological data streams; performing nonlinear dimensional compression mapping on continuously fluctuating parameters to generate dynamic mapping factors; calculating biochemical characteristics for discrete sampling parameters and triggering neutral degradation assignment to generate static modulation factors when anomalies occur; constructing a nonlinear decaying negative feedback term based on the previous blood volume state value and performing truncation processing to generate an adaptive saturation factor; fusing the above factors to update the unconstrained blood volume; and using numerical pinch intervals for physical boundary constraints to obtain the relative blood volume monitoring value. This application has the advantages of improving the system's anti-interference capability in high-frequency data sampling environments, blocking the propagation of missing data, and effectively suppressing system computational divergence and numerical oscillations.

The application discloses a large-scale data analysis method based on a DBSCAN algorithm, which comprises the following steps: step one, calculating an adjacency list through ring communication; step two, locally preliminarily determining a temporary cluster number of a core point; step three, alternately broadcasting a synchronous core point cluster number; and step four, updating a boundary point cluster number through reverse ring communication. The application can well solve the efficiency problem of the DBSCAN algorithm in large-scale data clustering analysis, ingeniously decomposes the calculation step dependence problem of the recursive calculation process of the DBSCAN algorithm, makes the algorithm applicable to a distributed execution model, makes the work load of different nodes in a cluster as balanced as possible, greatly reduces the communication amount in distributed calculation, and uses a non-blocking communication mode to make the calculation and communication overlap, reduces idle waiting time, fully utilizes the memory of a calculation node, and makes the massive data clustering analysis on the cluster become a reality.

The application discloses a user residence point identification method and device, electronic equipment and storage medium. Target signaling is added to each user in first user signaling data obtained, and second user signaling data is obtained. The target signaling is signaling with maximum time stamp for start time and end time and invalid values for longitude and latitude. Recursive calculation is performed on the second user signaling data through a recursive algorithm. A plurality of residence segments corresponding to each user are identified according to a distance threshold and a time threshold. The signaling length of each signaling in each residence segment is used to assign a corresponding weight to each signaling, and the offset calibration residence point of each residence segment is calculated. Finally, the offset calibration residence point of each residence segment and the second user signaling data are judged through the recursive algorithm, and the corresponding residence point of the plurality of residence segments of each user is output. Thus, the accuracy of identifying the residence state of the user can be improved.

The invention relates to the field of data analysis and intelligent decision, and particularly discloses a multi-level contribution deconstruction and aggregation method for an element system based on machine learning, which comprises the following steps: extracting element observation data, generating a standardized element feature vector, and forming a system feature matrix; training a relational machine learning model by using the system feature matrix and the overall efficiency data, and automatically generating a multi-level contribution deconstruction tree; selecting nodes in the contribution deconstruction tree for verification, and dynamically adjusting and optimizing the model and the tree structure according to the deviation between a predicted value and an actual value to obtain a final contribution deconstruction tree; based on the final contribution deconstruction tree, contribution value weighted aggregation is carried out from bottom to top, and the comprehensive contribution rate of each element is calculated through recursion. According to the method, the complex contribution relationship is automatically and objectively deconstructed through machine learning, the accuracy and adaptability of the model are ensured by means of verification of a closed loop, the quantitative contribution rate with a clear structure and high interpretability is finally output, and the reliability and decision support value of a calculation result are remarkably improved.

Circuits and methods for computing an order N polynomial include V decimation stages, each stage including respective multiply-and-accumulate circuitry. The multiply-and-accumulate circuitry in each stage k, in response to an input r-term and a plurality of input z-terms 0 through (Nk−1), generates output z-terms 0 through (Nk / 2−1) and an output r-term as a square of the input r-term. Each output z-term i is a sum of input z-term (2i+1) of the input z-terms and a product of input z-term 2i and the input r-term. The multiply-and-accumulate circuitry in decimation stages k for k≤(V−1) provides the output r-term and one or more output z-terms from decimation stage k as the input r-term and one or more input z-terms to the respective multiply-and-accumulate circuitry of decimation stage k+1. A recursive stage inputs from decimation stage V, the output r-term as a recursive r-term and the output z-terms as a-terms, and generates a polynomial output value z by a recursive evaluation of the recursive r-term, the a-terms, and a modulus, p.

This invention belongs to the field of carbon management technology, specifically relating to a product carbon footprint modeling method and system. Through a dynamic construction and optimization mechanism of node example networks, this invention significantly improves the modeling accuracy and computational efficiency of carbon emission data in complex production processes. Compared to traditional tree-based data models, this solution not only solves problems such as unclear data hierarchy semantics, low storage efficiency, and high complexity of recursive calculations, but also enhances the dynamic adjustment capability of the carbon emission tracking model, enabling it to better adapt to frequent changes in production processes. It achieves the verification of emission ratios of main and by-products and the integration of multi-dimensional carbon emission data, providing reliable data support for carbon management decisions. Simultaneously, through a dynamic scheduling mechanism of computational resources among heterogeneous subgraphs, it effectively improves stability and response speed in large-scale carbon footprint calculation scenarios, significantly shortening the carbon footprint assessment cycle while ensuring computational accuracy.

The application discloses a storage pool creation method and device, equipment, medium and product, applied to the storage technical field, including: obtaining the number of fault domain units and the number of copies; when the number of fault domain units is greater than the number of copies, whether each fault domain unit meets the first preset capacity balance condition is judged based on the minimum effective capacity corresponding to each fault domain unit and the expected effective capacity; if the first preset capacity balance condition is not met, the target effective capacity is obtained by grouping and recursively calculating each fault domain unit, and whether the second preset capacity balance condition is met is judged based on the target effective capacity and the expected effective capacity; in the case of not meeting the second preset capacity balance condition, the main pool is created based on the target effective capacity, and the auxiliary pool is created based on the remaining capacity. In this way, the utilization rate of storage resources can be improved.

The application relates to the technical field of database operation and maintenance, in particular to an automatic operation and maintenance method for an audit system database, which comprises the following steps: periodically collecting the inventory of each level of an LSM-Tree and a threshold value, and calculating a level capacity occupation ratio; using an eBPF technology to count the background I / O time consumption and flow in a sliding window, and calculating the background disk writing unit time consumption; performing recursive calculation in the reverse order direction by using the level capacity occupation ratio, so as to obtain a full-link logical write multiple; analyzing the coupling relationship between the level capacity occupation ratios of adjacent levels, determining a multi-layer concurrent congestion coefficient, obtaining a unit data estimated occupation time length, reversely deducing a front-end allowed write rate, and combining a token bucket algorithm to smooth the flow limiting of the front-end audit write request. The application aims to realize the early perception of the internal backlog trend of the LSM-Tree, and automatically perform predictive flow control, so as to optimize the operation and maintenance management of the audit system database.

The application belongs to the technical field of power electronic real-time simulation, and aims at the problems of insufficient simulation accuracy of traditional pure CPU simulation, limited simulation scale of FPGA simulation, communication bottleneck and poor flexibility of "CPU+FPGA" hybrid simulation, and proposes a power electronic converter small step real-time simulation method and system. The core is to build a multi-port equivalent model based on the switching function method or the improved node method, and to equivalent the nonlinear part of the converter into a controlled source and fuse the linear passive network; adopt the "CPU unique calculation + FPGA special communication" architecture, FPGA nanosecond level acquisition PWM pulse sequence transmission to CPU, CPU through the main step nesting small step recursive calculation, in the microsecond level main step, reproduce the hundred nanosecond level electrical dynamic. The matching special system deeply optimizes the CPU multi-core parallel ability, the communication cost is low, realizes the accuracy, the efficiency and the scale unification, the cost is low, the flexibility is high, can support the large-scale high switching frequency system high precision simulation, is applicable to the new energy grid connection, direct current transmission and other scenes.

This vehicle mass estimation method includes: detecting an acceleration generated in a vehicle in the front-rear direction; setting, on the basis of the state of the vehicle, a value of a forgetting factor for past data to be forgotten; varying the value of the forgetting factor according to the elapse of time; and estimating a vehicle mass by recursive computation based on an recursive least square method using the forgetting factor on the basis of a vehicle model including the vehicle mass, the driving force of the vehicle, and a detected value of the acceleration.

The invention relates to the technical field of image analysis, in particular to a facial occlusion image redrawing network method, which comprises the following steps: migrating the time sequence modeling advantage of RWKV to the field of image redrawing to realize semantic perception and identity consistency reconstruction of an occlusion area. Reconstructing a calculation normal form of a WKV attention mechanism and a token conversion layer, and decoupling recursive calculation based on time steps in RWKV into feature interaction of spatial dimensions, so that the feature interaction adapts to structural characteristics of a two-dimensional image; a window circulation WKV attention module is designed, multi-direction feature propagation is simulated through sub-window scanning and a bidirectional circulation mechanism, and local fine-grained texture and global structure constraint modeling are synchronously captured in a single-layer network; according to the method, a multi-scale token transfer layer is constructed, and the context dependency relationship of fine structures such as interbrow wrinkles and lip lines is re-calibrated and strengthened through cross-scale features, so that the problem of detail homogenization caused by occlusion is avoided, and the redrawing effect of a facial occlusion image is improved.

The invention provides a forward measurement normalization method and system for CCSDS131.0-B-3 standard Turbo decoding, electronic equipment and a storage medium, and the method and system are used for carrying out the systematic simulation analysis of the value change behavior of forward measurement Alpha of each state in the decoding process of a CCSDS131.0-B-3 standard Turbo code frame. Obtaining statistical characteristics of each state which becomes a forward measurement Alpha maximum value state in a forward recursion process; the method comprises the following steps: based on statistical characteristics, screening out a fixed state from all states of a CCSDS131.0-B-3 standard Turbo code grid diagram as a reference state for forward measurement Alpha normalization in the whole decoding process; when recursive calculation of forward measurement Alpha is carried out by a Log-MAP algorithm, for each moment t of a decoding process, a measurement value of a reference state at the current moment t is subtracted from non-normalized forward measurement values of all states at the current moment, so that normalization operation of the forward measurement Alpha of all states at the moment is completed.

The invention provides an electronic archive data processing method and device. The method comprises the following steps: performing content segmentation on to-be-processed electronic archive data according to content characteristics of the to-be-processed electronic archive data to obtain a plurality of data blocks; generating a corresponding hash value for each data block by adopting a hash algorithm; inquiring whether a hash value exists in a global fingerprint index database or not; if so, establishing a logic reference relationship pointing to the existing data block; if the hash value does not exist, storing the data block corresponding to the hash value, and registering the hash value into the global fingerprint index database; and on the basis of the hash values of the plurality of data blocks, constructing a Merkle tree-shaped data structure taking the hash values as nodes, and generating a root hash representing the complete state of the file through recursive calculation. According to the method, an endogenous and efficient data representation and integrity verification mechanism is established for the electronic archive through a core method flow of blocking, duplicate removal and Merkle tree construction.

The embodiment of the invention provides an index weight evaluation method and a related device. The method comprises the steps that original text data is acquired, the original text data comprises index system data and fact data, and the index system data comprises names of different indexes and hierarchical relations between the indexes; preprocessing the original text data to obtain target text data; performing semantic vectorization processing on the target text data to obtain a semantic vector; performing similarity calculation on the semantic vectors to obtain hierarchical semantic similarity and data semantic association degree; according to a preset fusion strategy, fusing the hierarchical semantic similarity and the data semantic association degree into an initial weight of each index; performing normalization processing on the initial weight to obtain a relative weight; and recursively calculating the relative weight of each level of indexes in the index system data, and generating an index weight evaluation result. On the basis, the comprehensiveness and reliability of index weight evaluation can be improved.

The invention relates to a computer-implemented method, a device, a computer program, and a memory for controlling pressure or clamping force in electromechanically actuated systems. It comprises: receiving a differential pressure by an inverse model module based on a desired or target pressure; calculating an unlimited volume flow by the inverse model module based on the input differential pressure; inputting the unlimited volume flow to a limiter module configured to generate a limited adjustable volume flow; inputting the limited adjustable volume flow to a system model module configured to calculate an adjustable pressure; and using the model-based, recursively calculated target pressure in subsequent calculation cycles, including in a closed-loop control system.

The invention provides a numerical value coding method for high-dimensional spherical surface uniform distribution based on an irrational number angle. The method comprises the following steps: obtaining a preset irrational number list composed of N-1 linearly independent irrational numbers and an integer coding value; then multiplying each irrational number by the integer coded value and carrying out modulo processing to generate an N-1-dimensional angle sequence; initializing an all-one vector based on the sequence, and recursively calculating first N-1 components and a final component according to a conversion rule from spherical coordinates to Cartesian coordinates to form a non-normalized initial vector; strict normalization is realized through Euclidean norm calculation and component scaling. According to the method, ergodicity and aperiodicity of irrational numbers are utilized, random sampling is replaced by deterministic calculation, under the condition that unit spherical surface constraint is guaranteed, strictly and uniformly distributed coding vectors are efficiently generated on a unit spherical surface of any dimension, the problem that calculation efficiency is reduced due to dimension disasters is effectively solved, and the method is suitable for large-scale popularization and application. And a feature representation basis with geometric consistency is provided for a machine learning model.

The invention provides a sequence processing method and device based on attention division, equipment and a storage medium, relates to the technical field of data processing, in particular to the fields of large language models, multi-modal large models and the like, and can be used for application scenes such as complex reasoning and decision making. According to the implementation scheme, an attention matrix is generated by utilizing a multi-head attention mechanism according to vector representation of a to-be-processed sequence; according to the complexity of the attention matrix, dividing the plurality of attention heads into a first type of attention heads and a second type of attention heads; based on the vector representation, determining an initial attention vector by using the first type of attention head and the second type of attention head; based on the initial attention vector, recursive calculation is carried out by using a second type of attention heads until a preset cycle step number is reached, and an attention vector corresponding to the attention layer is obtained; and generating a processing result of the to-be-processed sequence according to the attention vector. According to the scheme, the calculation redundancy can be remarkably reduced, and the model reasoning ability is improved under the same parameter scale.

The invention relates to the field of electric digital data processing, in particular to a self-adaptive harmonic suppression method and system for a motor home inverter. According to the method, an inverter output voltage error signal is obtained through sampling, and the signal is processed to extract a first characteristic quantity reflecting the waveform local smoothness of the signal. And based on the first characteristic quantity, dynamically generating the time-varying forgetting factor through a preset monotone decreasing mapping relation. The forgetting factor is used for conducting recursive calculation on the error signal or the function square value of the error signal to obtain an elastic energy value, and the calculation inertia of the elastic energy value is enhanced along with increase of the forgetting factor. A normalized relative deviation between the elastic energy value and a reference energy value obtained by a moving average method is then calculated. And finally, according to the relative deviation degree, reverse adjustment is carried out on the gain of the harmonic suppression controller: the gain is reduced when the deviation degree is increased, and the gain is increased when the deviation degree is reduced. The method can effectively adapt to complex working conditions, and the system stability is improved.

The invention belongs to the technical field of artificial neural networks and brain-like computing, discloses a long-time sequence classification method and device based on a spiking neural network and a storage medium, and aims to solve the problems that an existing spiking neural network depends on recursive computing in a time sequence modeling process, time dimensions are difficult to be parallel, and training efficiency is low. Relates to a spiking neural network calculation structure based on multi-branch adaptive dilated convolution, and aims to realize high-efficiency parallel modeling of a long time sequence dependency relationship while maintaining the event-driven calculation characteristic of a spiking neural network, thereby improving the training efficiency and time sequence modeling capability of a model.

A technique uses linear prediction to determine the location of spurious content in a digital phase-locked loop and suppresses the spurious content from propagating to the clock output. In at least one embodiment, the technique implements an iterative (e.g., recursive) computation.