599 results about "Directed acyclic graph" patented technology

The invention relates to the technical field of intelligent manufacturing and production management, in particular to a self-adaptive production scheduling system based on artificial intelligence, which comprises a data acquisition and reference construction module for analyzing process data to construct a directed acyclic graph representing a non-interference state as a reference map; the theoretical disturbance simulation module is used for converting the interference rule into a graph change instruction, generating a theoretical damaged state graph and obtaining a theoretical difference feature vector; the theoretical difference feature vector comprises, but is not limited to, a vector form obtained after a difference matrix is expanded according to rows or columns in terms of mathematical representation; the real deviation extraction module is used for collecting real-time state data to construct a real-time operation state diagram and calculating a real difference feature vector; a double-domain coupling decision module; an adaptive scheduling execution module; according to the method, the causal relationship is verified by comparing the form of theoretical deduction and actual observation, non-systematic noise is effectively filtered, and accurate response to real faults is realized while the stability of the production rhythm is maintained.

The invention provides an underground equipment fault real-time diagnosis method and system based on edge calculation, and relates to the technical field of coal mine safety production, and the method comprises the steps: collecting multi-modal data through a distributed sensor network, extracting multi-scale time sequence features, projecting the features to a Lie group manifold space, constructing a coupling mapping relation matrix, obtaining fusion features, and carrying out the real-time diagnosis of an underground equipment fault; and constructing a causal directed acyclic graph based on a topological connection relationship and a Granger causal coefficient, executing Bayesian probabilistic reasoning, determining an execution strategy in combination with entropy similarity matching, and performing deep time-frequency analysis and causal chain verification. High-precision real-time diagnosis of equipment faults in an underground complex environment is realized, and the fault early warning accuracy is improved.

The invention relates to an unmanned aerial vehicle path planning method and system based on GNN and high-order security constraints. The method comprises the steps that a navigation scene where an unmanned aerial vehicle is located is represented as a heterogeneous directed graph, message passing and feature updating are conducted through the GNN, a risk-aware attention mechanism is introduced, and an interpretable decision result is output; a differentiable HoCBF-QP optimization layer is introduced, an original control command output by a strategy network is used as input, quadratic programming with high-order control barrier function constraints is solved online, minimum-amplitude safety correction is carried out on the control command, and an actuator control command meeting safety constraints is output; starting a HoCBF safety shield during operation so as to strictly ensure that all safety constraints are met before execution; a calculation task of the whole control cycle is modeled into a directed acyclic graph form, and parallel execution is carried out on heterogeneous multiple cores by utilizing a real-time scheduling strategy. The problem that unmanned aerial vehicle navigation control is not effective and unified in three aspects of structure, safety and scheduling is solved.

The invention discloses a multi-modal data and dynamic knowledge graph fusion method, relates to the technical field of knowledge graphs, and solves the technical problems that multi-modal data weight evaluation is single in dimension and lacks a data source, time and scene collaborative quantization mechanism. In combination with data source authority and time freshness, it is ensured that data value evaluation better meets service requirements, dimension weight coefficients can be flexibly adjusted according to scenes, different from fixed coefficients in the prior art, requirements of different fields are met, data full-link metadata is stored through a directed acyclic graph, conflict sources can be rapidly traced, and the data value evaluation efficiency is improved. And automatic classification and targeted correction of conflict types are realized based on a predefined rule, and the error rate of conflict correction is reduced.

The invention relates to an enterprise data operation multi-source data virtualization access and index automatic management and control method and system, belongs to the technical field of computers, and aims to solve the problems of index definition fragmentation, frequent logic conflicts, dependency static management and verification rule rigid coupling in the prior art. Comprising the following steps: constructing a unified virtualized data access layer to integrate heterogeneous data sources; the indexes are disassembled into semantic atoms capable of being managed in a versioning mode and registered to an atom knowledge base; dynamically constructing an index dependency relation directed acyclic graph based on semantic atom combination; dynamically binding the data quality rules with semantic atoms; and analyzing an index request through a virtualized query engine, reversely traversing atlas aggregation dependency and rules, generating and executing local query, and automatically checking result data quality at the same time. According to the method, unified index definition, dependence on automatic conduction, quality embedded verification and multi-source transparent access are realized, and the data consistency, the treatment efficiency and the system flexibility are remarkably improved.

The invention provides a DAG arrangement and execution method and system based on operator dynamic management, and the method comprises the steps: constructing an initial directed acyclic graph DAG, dynamically loading operator plug-ins registered in the initial DAG, obtaining context information during operation, and storing the context information to a context container; dynamically selecting a next execution path based on the real-time data in the context container, generating an executable operator queue according to the updated DAG structure, and monitoring the execution state of the operator node in real time; and dynamically modifying the DAG structure according to a trigger condition in the execution process, and outputting a final result after the DAG is executed. According to the DAG arrangement and execution method and system based on operator dynamic management, flexible configuration of the task process, dynamic scheduling of the operator level and intelligent decision making during operation are achieved, and therefore the universality, adaptability and execution efficiency of the system are remarkably improved.

The invention provides a multi-modal neural causal inference micro-service fault positioning method and system, and the method comprises the steps: accessing observability data in a service operation process, and representing the tracking information of each request as a directed acyclic graph of a multi-modal feature; performing multi-modal feature coding and graph self-coding anomaly detection on the calling graph, and identifying an abnormal node through a reconstruction error; based on service topology prior, learning a sparse causal relationship graph between services by adopting a multi-scale neural causal inference method; calculating a node root cause score according to the causal relationship graph and the abnormal score, and executing causal path search to generate a fault propagation path; and marking the potential root cause according to the path weight of the propagation graph and the node popularity, and outputting a visual diagnosis result. According to the method, the system operation state is comprehensively described by fusing three kinds of micro-service system multi-modal data of logs, indexes and Trace in the micro-service system, and the structure-perceived causal diagram is constructed, so that accurate and explainable root cause positioning is realized.

The invention discloses a customer complaint work order automatic circulation system and method based on composite intention disassembly. The system receives and preprocesses the multi-mode customer complaint data and initializes a service circulation context; business intention identification and business key slot position extraction are carried out; disassembling the composite business intention, and constructing an ordered work order execution queue of a directed acyclic graph structure through cyclic dependence verification; evaluating the work order circulation action with the maximum expected cumulative income by using a reinforcement learning model; when business knowledge query is involved, compliance evidence fragments are retrieved, sorted and output; and generating a candidate reply and a business operation instruction by the fine-tuned large language model, outputting the candidate reply after business compliance verification, and calling an API (Application Program Interface) of an underlying business system to execute entity business handling operation in a cross-system manner. The problems of data interaction and state synchronization among multi-source heterogeneous systems are solved, the automatic execution efficiency of the composite work order at the bottom layer node is improved, and the system unauthorized and dirty data risks caused by an uncertain instruction are effectively avoided.

The invention discloses a multi-dimensional data asynchronous calculation method and system based on a dynamic dependency graph, and relates to the field of data processing. The method comprises the following steps: constructing a metadata, logic and instance three-layer separation storage model; analyzing the reference relationship to construct a directed acyclic graph, and generating a calculation priority; monitoring data change, and executing asynchronous serialization calculation through a message queue based on priority; the associated document is automatically updated based on anchor mapping. According to the method, the calculation deadlock and the performance bottleneck of large-scale data in the Web environment are solved, logic decoupling and dynamic expansion are realized, the final consistency of the data is guaranteed, and the high-concurrency throughput and the stability are remarkably improved.

The invention discloses an underwater acoustic processing CPU-GPU dynamic load balancing method based on a task flow model, and belongs to the technical field of underwater acoustic processing. The method comprises the following steps: deconstructing an underwater acoustic processing application into a task flow model represented by a directed acyclic graph; establishing a feature portrait including calculation complexity, parallelism and data throughput for each task node, and constructing a cost prediction model; the CPU / GPU utilization rate and the data transmission performance are monitored in real time; a processor is distributed to each task node by using a dynamic programming algorithm in combination with a cost prediction model and a real-time system state with the goal of minimizing the total task flow completion time; and dynamically scheduling tasks through a central scheduler according to a decision result, and periodically updating a strategy. According to the method, the defects that static task division lacks adaptability and neglects task dependence and communication overhead are overcome, dynamic and efficient utilization of CPU and GPU resources is achieved, and the efficiency and real-time performance of underwater sound processing are remarkably improved.

The invention relates to a query optimization method and device based on agent planning, electronic equipment and a medium. The method comprises the following steps: pre-retrieving a target tree structure index based on a user query request to obtain a pre-retrieval result set; generating a retrieval-reasoning double-stage directed acyclic graph task chain comprising a task type, a retrieval / reasoning target and a dependency relationship; executing each task according to a topological sequence, and obtaining a target text fragment through query rewriting and expansion; and performing multi-hop logical reasoning in combination with the task chain and the target text fragment, and outputting a final answer. Therefore, dynamic task planning of environment perception is carried out by constructing a retrieval-reasoning dual-stage directed acyclic graph task chain and combining a knowledge base tree structure index; the problems that when an existing retrieval enhancement generation system processes a multi-hop question and answer task, logic association understanding ability among retrieval fragments is insufficient, semantic matching precision is low, and a reasoning path is prone to deviation are solved, and the retrieval recall rate and the answer accuracy under a multi-hop question and answer scene are improved.

The invention relates to the technical field of block chain technology and artificial intelligence collaboration, and discloses a verifiable multi-agent collaboration method based on smart contract driving, comprising the following steps: task planning generates a directed acyclic graph and a zero-knowledge proof circuit template, submits the directed acyclic graph and the zero-knowledge proof circuit template to a planning contract, and deploys an exclusive zero-knowledge proof contract; the AI agent collaboratively executes the subtasks, and zero-knowledge proof is generated and submitted through a Veckrii auction tool; zero-knowledge proof contract chain verification is carried out, and after verification is passed, a signature is received and multi-layer gating verification is executed; and the governance contract carries out pledge management and cost distribution on the agents, the providers and the experts. According to the method, zero-knowledge proof and a trusted execution environment are combined, and an under-chain execution behavior is converted into an encryption certificate capable of being verified on a chain, so that the problem that the execution process cannot be verified due to the lack of distributed trusted evidence storage in a traditional centralized framework is effectively solved; and after each subtask is completed, double signature and multi-layer gating verification are forcibly executed.

The invention discloses a financial monitoring system based on real-time data processing and intelligent analysis, and relates to the technical field of financial information, and the system comprises a bill collection module which employs a configuration API interface, carries high-concurrency data streams through a distributed message queue, carries out the real-time connection of an upstream business system and a financial system, and receives bill detail data; and the analysis and check task arrangement module manages the task dependency relationship based on a directed acyclic graph model, and arranges a data pulling task, a bill general task and a bill check task according to a preset rule. According to the financial monitoring system provided by the invention, the automation and intelligence level of financial management is remarkably improved by integrating real-time data processing and intelligent analysis technologies, the system adopts a configured API interface and a distributed message queue, efficient butt joint with an upstream business system and a financial system is realized, and the financial management efficiency is improved. The real-time performance and accuracy of data receiving are ensured, and the data processing flow is more flexible and controllable.

The invention relates to a traffic simulation agent system construction method based on a large model, and the method comprises the steps: constructing a simulation tool library, and achieving the precise evaluation and continuous optimization of a simulation result through the fusion of multi-source heterogeneous traffic data, the construction of standardized input, and the establishment of a quantitative evaluation system. A large language model is utilized to understand a natural language instruction of a user, tasks are intelligently disassembled, an execution process is planned, dependence management and parallel scheduling are carried out in combination with a directed acyclic graph, and professional tools are driven to automatically execute. And performing evaluation, problem diagnosis and adaptive re-planning on an execution result through a large language model reflection mechanism to form an understanding-planning-execution-reflection closed loop. According to the method, the problems of how to assist a user to interact with a traffic system by utilizing an agent technology driven by a large language model, reducing the technical threshold of traffic simulation software use and saving time cost and labor cost are solved, the traffic simulation use threshold is reduced, the automation and intelligence level is improved, and efficient and accurate traffic system interaction and optimization are realized.

The invention relates to a remote sensing image data processing method and system based on a Spark architecture, and belongs to the technical field of remote sensing image data processing. The method comprises the following steps: converting a collected original file into a COG format, constructing a hybrid index through a spatial dimension reduction technology, extracting metadata and establishing a metadatabase; based on the hybrid index, quickly positioning a target tile through spatial query, reading a data block, and constructing a TileLayerRDD of a spatial partition in a Spark cluster; and dynamically assembling and calculating the directed acyclic graph according to the algorithm warehouse template, and submitting the directed acyclic graph to a Spark cluster for distributed memory calculation. According to the method, rapid retrieval is achieved through the COG format and the mixed index, pre-slicing is replaced with Spark distributed computing, the storage pressure is greatly reduced, and on-demand real-time analysis and high-concurrency online service of mass remote sensing data are finally achieved through micro-service and elastic resource scheduling.

The invention discloses an Agent task planning and workflow generation method driven by knowledge in the semiconductor field, and relates to the cross technical field of artificial intelligence and semiconductor manufacturing. According to the method, in order to solve the problems that in the prior art, a semiconductor data analysis task lacks domain knowledge support, workflow generation stability is poor, and execution links are uncontrollable, a Task-Planner and Executor-Planner double-layer framework is adopted, and accurate matching from user requirements to domain subtasks is achieved by constructing a semiconductor domain task knowledge base; a structured prompt mechanism is designed, and a directed acyclic graph workflow with complete logic is generated in combination with the fine-tuned large language model; error localization processing is introduced, and reliable execution of the workflow is guaranteed; the method can automatically complete the planning and execution of complex tasks such as yield analysis and fault diagnosis in semiconductor wafer manufacturing, packaging testing and equipment operation and maintenance, improves the data analysis efficiency and reliability, and is suitable for a high-precision semiconductor manufacturing scene.

The invention discloses a finance and taxation data analysis method, device and equipment based on a directed acyclic graph and an intelligent agent and a medium, and relates to the technical field of finance and taxation inspection, and the method comprises the steps: constructing a target directed acyclic graph which takes knowledge blocks as nodes and dependencies as directed edges based on a historical finance and taxation data set and natural language description of the knowledge blocks; performing hierarchical division according to the in-degree of each knowledge block in the graph, and generating a reference task plan by using a target tool list based on each knowledge block in different hierarchies, a thinking chain guide instruction and a cue word generated by an example; controlling the intelligent agent to execute the planning, and packaging the reference task planning according to the obtained execution result set and the log to obtain a basic task planning template; utilizing the current finance and tax data and the current to-be-analyzed demand to finely adjust the context parameters and the analysis range in the template to obtain a target task plan; and analyzing the current finance and tax data set based on the plan to generate a target analysis result. And the finance and tax data analysis consistency and reliability are improved.

An energy-conscious workflow planning system based on LLM with dynamic resource allocation, consisting of: a workflow input interface configured to receive workflow-directed acyclic graphs (DAGs), energy budget constraints, system performance constraints, and natural language requests from human operators; a large language model (LLM) logic module connected to the workflow input interface and configured to analyze the workflow specifications and system constraints in natural language, generate energy-conscious planning recommendations based on the analyzed workflow specifications, and provide explainable planning rationales in natural language; a reinforcement learning-based scheduling unit connected to the LLM reasoning module and configured to: receive scheduling recommendations from the LLM reasoning agent, fine-tune task-resource assignments by dynamically adapting to runtime variations, and perform online resource redistribution under runtime variability; an energy monitoring unit configured to: continuously monitor CPU and GPU utilization in heterogeneous clusters, track power consumption and thermal limits per node, and generate energy profiles for system components; a multi-objective optimization engine configured to: perform a Pareto-optimal scheduling analysis that balances energy consumption, lead time and reliability, apply statistical and AI-supported trade-off analyses and ensure optimal resource allocation based on Pareto frontier analysis; a dynamic resource allocation unit configured to: use predictive models that incorporate LLM inferences and feedback from reinforcement learning, reassign tasks between nodes and clusters while minimizing energy consumption and improving system throughput based on the predictive models; a performance optimization module configured to optimize scheduling decisions using multi-criteria optimization analysis; and a user interface that allows human operators to override and refine planning strategies in real time based on verifiable planning reasons.

The invention relates to the technical field of big data management, in particular to a data tracing method based on big data mining, which comprises the following steps of: acquiring an unstructured operation log, analyzing the unstructured operation log into a structured event record, and associating to generate a heterogeneous operation session set; performing access mode association analysis on the session set to extract a read-write path mode, and combining association strength and a time decay factor to determine a dependency weight and construct a blood relationship map; eliminating a loop of the graph to generate a directed acyclic graph, constructing a spanning tree serving as a retrieval trunk and a node index coding interval, and mapping a cross-branch associated edge to a bitmap index to generate a probability skeleton tree index; and positioning a trunk path by using the interval, calling a bitmap to perform multi-path verification, and outputting a traceability path and confidence. A probability consanguinity map and a probability skeleton tree index are constructed through mining logs, and data consanguinity reconstruction and large-scale node low-delay retrieval in the heterogeneous black box environment are achieved.

The invention discloses a multi-agent cooperative task method and device based on a hierarchical task mechanism, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the hierarchical abstraction based on a task space of a target domain, obtaining a plurality of hierarchies, configuring a corresponding agent and a tool for each hierarchies, constructing a state management mechanism to manage intermediate data among a plurality of hierarchies; in response to target information input by a user, decomposing and planning a target task by using a plurality of mutually associated hierarchies to obtain a directed acyclic graph of the target task; calling and executing the subtasks corresponding to the nodes from bottom to top based on the directed acyclic graph so as to complete the target task; and integrating a completion result of the target task and then rendering the result to a user interface. The problem that an existing intelligent agent system is difficult to meet the requirements of complex fields with highly structured and hierarchical features is solved. A more efficient, accurate and robust scheme can be provided for complex tasks with highly structured and hierarchical features.

The invention provides a data decoupling centralized query method and system based on MQ message triggering, and relates to the technical field of data query, and the method comprises the steps of receiving a message queue data change message, constructing a field, dividing check nodes by depending on a directed acyclic graph, carrying out data consistency verification, and pushing a transaction identifier to carry out concurrent operation scheduling. And collecting hotspot information by adopting a Bloom filter to adjust data fragmentation and cache distribution, generating a distributed query routing table, and distributing query requests according to the routing table. According to the method, the data consistency is improved, and the query performance is optimized.

The invention discloses a multi-model reasoning method and device based on a graph structure, a medium and a program product, and the method comprises the steps: querying a task target node in a graph database, starting from the task target node, traversing the graph database according to a corresponding driving model, and carrying out reasoning path extension, each directed acyclic graph is obtained by mapping metadata of the registered analysis model; generating a task topology execution graph according to the model nodes involved in the expanded reasoning path and the corresponding driving model; and according to the task topology execution graph, actual execution of each model node is scheduled in sequence, and an interpretation result corresponding to the task target node is obtained. According to the method, the model attribution relation is explicitly analyzed, and the reasoning path is optimized by utilizing graph calculation, so that the reasoning execution efficiency is improved, and the intelligence, performance and interpretability of multi-model collaborative analysis are improved.

The invention relates to a deterministic service-oriented resource scheduling method, apparatus and device, a medium and a program product. The method comprises the following steps: modeling a deterministic service into a directed acyclic graph; therefore, the processing flow of the deterministic service can be represented by the directed acyclic graph, so that the intelligent agent can identify the execution overhead of each task on the computing node according to the directed acyclic graph, and a basis is provided for subsequent scheduling decisions. Determining a real-time state of each computing node in the heterogeneous computing environment; therefore, the current data processing capacity of each computing node in the complex heterogeneous computing environment can be mastered in real time, and the task load of each computing node can be balanced conveniently. And based on the directed acyclic graph corresponding to the deterministic service and the real-time state of each computing node, outputting a scheduling decision of the deterministic service through the trained agent. Therefore, computing and communication resources can be dispatched cooperatively under a unified converged network architecture, so that the computing and communication resources can adapt to a complex and changeable network environment, and an optimal dispatching decision is output.

The invention discloses a cross-modal causal coupling water supply network cascade failure deduction method and system, and relates to the field of safety prevention and control of urban water supply networks. The method comprises the following steps: firstly, constructing a water supply network heterogeneous graph, abstracting physical pipe sections as core nodes, associating the pipe sections as heterogeneous edges, and obtaining static and dynamic attributes of the heterogeneous edges; multi-source heterogeneous data space-time alignment is completed through an exclusive scale adaptive coding layer, and a standardized feature matrix is generated; taking the cross-modal causal directed acyclic graph as a hard constraint, and generating a dynamic coupling adjacency matrix through a cross-modal graph attention layer; dynamic weighted coupling of multi-modal characteristics is realized through a space-time dynamic gating unit; finally, cascade failure evolution is simulated based on the coupling characteristics and the adjacent matrix, critical inflection points are recognized, and graded early warning signals are output. The method solves the problems of modeling distortion, false correlation and inflection-point-free pre-judgment in the traditional technology, greatly improves the deduction precision and the engineering credibility, and is suitable for safety prevention and control and emergency early warning of the urban water supply pipe network.

The invention provides an industry vertical data fine tuning method and device based on a large model, and the method comprises the steps: obtaining an operation order text data flow, carrying out the preliminary structural processing of each operation order text, extracting an operation action, an execution object and a constraint condition, and forming text data carrying an initial label; determining an operation process directed acyclic graph and an emergency operation process network according to the fine tuning corpus of the text classification model, and taking the operation process directed acyclic graph and the emergency operation process network as structured training samples for fine tuning of the text classification model; determining a mixed training batch of fine tuning of the text classification model according to the structured training sample of fine tuning of the text classification model, and setting a dynamic balance threshold value of the standard operation sample and the emergency operation sample; and after fine tuning of the text classification model is completed and deployment is carried out, if the operation ticket text to be examined belongs to emergency operation, parallel or conditional branch logic reasoning processing is carried out based on the structure of the emergency operation process network, and a structured examination report is output.

The application provides a smart question and index management engine and system based on dynamic reward optimization, relates to the technical field of data learning, and is used for enterprise index analysis, attribution diagnosis and decision support. The system sets an index management layer to uniformly manage index caliber, calculation logic, blood relationship, credibility score and version information by means of a dynamic knowledge graph, uniformly implements semantic constraints on a multi-agent analysis process, and guarantees index consistency and traceability. The system also establishes a cause-effect cognition module, generates and corrects a business index cause-effect directed acyclic graph based on time series data and in combination with expert priori, realizes root cause positioning and counterfactual simulation, and answers "why change" and "what if intervention". A multi-agent collaborative analysis core is responsible for natural language intention analysis, index compliance verification, automatic data acquisition, cause-effect inference, narrative generation and chart presentation, and calculates a multi-target compound reward value based on user feedback and interaction behavior, and adaptively adjusts output; user correction write-back forms a closed-loop learning.

The invention belongs to the technical field of order fulfillment process monitoring, and discloses an order fulfillment process delay early warning method and system based on multi-node state acquisition, and the method comprises the steps: collecting the dynamic state data of multiple nodes in a fulfillment process in real time; matching judgment is carried out through the node rule base, primary early warning is triggered, and an early warning context data set is constructed; inputting the data set into a pre-trained delay risk prediction model, and outputting an overall risk probability and a key influence node identifier; issuing a grading early warning signal according to the risk probability; carrying out interpretable root cause positioning through a root cause analysis rule chain based on directed acyclic graph representation; incremental learning is carried out on the prediction model based on early warning feedback data to form an optimized closed loop; the system correspondingly comprises a data acquisition module, a rule judgment module, a risk prediction module, an early warning release module, a root cause positioning module, a feedback learning module and the like. According to the invention, the early-stage, accurate and automatic early warning and root cause analysis of the full-process risk of order fulfillment are realized.

A system and method include learning a topological order of a plurality of variables in a directed acyclic graph based on real data, computing parameter estimate values corresponding to the real data, computing error values based on the real data, the topological order, and the parameter estimate values, generating simulated data from the parameter estimate values and the error values, such that simulated variables in the simulated data preserve a causal relationship between variables in the real data, and the simulated variables in the simulated data preserve a correlation relationship between the variables in the real data, and reorganizing and outputting the simulated data based on the topological order.

The invention discloses a large-model multi-agent-oriented video memory management method and system and a storage medium, and the method comprises the following steps: S1, receiving a task instruction, and decomposing a task into context fragments according to a fragment template by using a large model; s2, performing analysis and dependency relationship identification on the task by using a large model according to the task planning cue word, and constructing a task dependency graph; the task dependency graph is represented by a directed acyclic graph; s3, the large model predicts the life cycle state of the context segment represented by each context node based on the task dependency graph; s4, at the beginning of execution of the task, loading the related context fragment into the cache, and dynamically multiplexing and releasing the allocated cache according to the life cycle state of the context fragment; the task structure can be perceived, the context cache life cycle can be dynamically managed, and the video memory utilization rate is improved.

The invention relates to the technical field of data processing, and discloses a computing power server system layer optimization method and system oriented to a high-load scene. The method comprises the steps of collecting micro performance indexes such as the CPU instruction cycle number and the page table missing frequency, constructing a three-layer causal directed acyclic graph through Granger causal inspection, reducing a parameter search space based on bottleneck node reverse backtracking, generating an interpretable optimization decision with a causal path and contribution degree quantification, and carrying out optimization on the basis of the interpretable optimization decision. The problems that the performance bottleneck root cause cannot be accurately positioned and the optimization result lacks transparency in the prior art are solved. According to the method, bottleneck node reverse backtracking and parameter space pruning are performed based on the causal atlas, so that the problem of low optimization efficiency caused by incapability of accurately positioning a performance bottleneck root cause and blind exploration of a parameter space in the prior art is solved.