131 results about "Loop optimization" patented technology

The invention discloses a large model reasoning efficiency dynamic optimization and hardware sensing compression method. The method comprises the following five steps: S1, generating an input complexity signal representing calculation complexity; s2, synchronously monitoring a hardware resource index of the operation platform, and generating a hardware state signal reflecting a real-time load; s3, inputting the input complexity signal and the hardware state signal into a dynamic strategy selector, and generating a compression control signal through a pre-trained decision model; s4, according to the compression control signal, dynamic reconfiguration operation is executed on the large model weight and the activation value of the current reasoning task; and S5, performing reasoning calculation by using the reconfigured large model, and feeding back a hardware resource index to the step S2 in real time in the calculation process to form a closed-loop optimization link. According to the large model reasoning efficiency dynamic optimization and hardware perception compression method, the problems of low resource utilization rate, delay fluctuation and energy efficiency imbalance caused by a static compression method in dynamic input and heterogeneous hardware environments can be solved.

The invention discloses a memory access optimization method based on intelligent cache management, and relates to the technical field of computer storage. According to the method, spatial-temporal characteristics and semantic association data of memory access requests are collected in real time, a dynamic heat matrix is constructed, and a multi-dimensional access rule is fused to improve modeling precision. And inputting the dynamic popularity matrix into a hybrid prediction model, predicting a future access probability by using a time convolutional network, analyzing a competition relationship between data blocks through a graph attention network, generating a conflict pre-judgment weight and a corrected popularity ranking, and effectively reducing the cache jitter risk. On the basis of popularity ranking and conflict weight, a fragmented reinforcement learning algorithm is adopted to divide logic sub-regions, a differential reward function is designed to dynamically decide cache operation, and performance and energy efficiency requirements are balanced; and finally, through an online learning mechanism, combining real-time feedback to dynamically adjust a prediction model weight and strategy parameters, forming a closed-loop optimization link, and realizing adaptive stability under long-term load fluctuation.

The invention belongs to the technical field of resource scheduling, and particularly discloses a resource scheduling method and system based on reinforcement learning, and the method comprises the steps: describing the dependency and conflict relation between tasks through constructing a task causal relation graph which can be dynamically updated; constructing a state space and an action space based on a current task state and a causal relationship, training an intelligent agent by adopting a reinforcement learning model in combination with a multi-target reward function, and generating a scheduling and resource allocation strategy; linear programming and heuristic joint resource allocation are carried out under strategy guidance, and task priorities and causal relationships are dynamically adjusted in combination with task states; by collecting execution data and analyzing strategy deviation, a model structure and parameters are further adjusted, and a complete closed-loop optimization process is formed. According to the method, task priority dynamic adjustment, resource allocation strategy self-adaptive updating and scheduling process closed-loop optimization are realized, and the method is suitable for complex project management scenes under multi-task and multi-resource constraints.

The invention provides a legacy system-oriented large-model-driven API document automatic generation system, belongs to the crossing field of artificial intelligence and software development, and provides a multi-modal data fusion and closed-loop verification mechanism aiming at the defects of a traditional API document generation method in the aspects of semantic comprehension, dynamic context capture and multi-technology stack adaptation. A code static feature and a dynamic track during operation are analyzed through a multi-source data acquisition module, and an interface semantic feature is extracted in combination with a field self-adaptive large model of a semantic enhancement analysis module; deducing an implicit service rule by fusing static / dynamic characteristics through a graph neural network, and generating a standardized document conforming to an OpenAPI specification through a parameterized template generative adversarial network (PT-GAN); and finally, performing three-level verification and closed-loop optimization through a sandbox environment. The method supports a heterogeneous system of 16 programming languages such as Java / C + + / Python, interface version changes can be automatically recognized, document patches are generated, the problems of missing and outdated system documents and low maintenance efficiency are solved, and maintainability and integration efficiency of enterprise-level systems are remarkably improved.

The invention discloses an enterprise production management method based on digital twinning and workflow simulation, and particularly relates to the technical field of industrial internet and intelligent manufacturing, and the method comprises the steps: constructing a physical production system digital twinning body and business process workflow model, and carrying out the dynamic association through a model fusion engine to form an integrated digital twinning model; real-time event driving is used for deducing and simulating a future production process, and bottleneck and conflict prediction is output; based on the prediction result, utilizing a multi-objective optimization engine to generate a plurality of alternative scheduling schemes; performing parallel simulation quantitative evaluation on the KPI of each scheme, selecting an optimal scheme, analyzing the optimal scheme into a control instruction, and issuing and executing the control instruction; and model self-correction and closed-loop optimization are realized through real-time monitoring and feedback. According to the method, the problem of service and physical state disjunction caused by digital twinning and workflow independence is solved, the whole-process closed-loop management from prediction to execution is realized, and the production self-adaption and intelligent level is improved.

The invention discloses a contract review closed-loop optimization method and system based on user weak feedback, and the key points of the technical scheme are that the system comprises a user input interface; a contract analysis module; a review logic initialization module; an examination model module; an error history retrieval module; a result correction module; a user feedback storage module; a closed loop optimization module; according to the method, the accuracy and efficiency of contract review are improved through an intelligent review process, risk analysis is carried out on the contract by automatically analyzing the contract text, matching the review list and applying dynamically optimized review logic in combination with a large language model, and the system can identify potential problems and improve the contract review efficiency. According to the method, the review logic can be continuously adjusted through weak feedback of the user to form closed-loop optimization, the review rule can be flexibly adjusted according to contract characteristics, industry requirements and regulation changes, so that more accurate and personalized review services are provided for the user, and the review effect can be continuously improved in the process of continuously accumulating the feedback.

The invention discloses an aviation airborne software test case intelligent generation and closed-loop optimization method, which comprises the following steps of: knowledge graph construction: analyzing a DO-178C standard document and a related field document, extracting entities and relationships defined in the DO-178C standard document and the related field document, and constructing a field knowledge graph fused with DO-178C standard knowledge; initial test case generation: based on the domain knowledge graph, combining a static analysis result of the source code of the tested airborne software, and utilizing a large language model to drive and generate an initial test case set; and closed-loop iterative optimization: executing the test case, evaluating whether the structural coverage rate reaches the standard or not, automatically identifying uncovered codes when the structural coverage rate does not reach the standard, generating a supplementary test case for iterative optimization, and outputting a final test case set until a coverage rate target corresponding to the software security level is met. According to the method, the test quality and efficiency of the aviation airborne software can be improved.

The invention discloses a hybrid heterogeneous cloud workflow scheduling method based on reinforcement learning, and belongs to the technical field of cloud computing. The method comprises the following steps: aiming at a cloud workflow scheduling problem, by taking minimization of completion time as a target and taking cost and resources as constraints, a three-dimensional collaborative constraint model is constructed, and the cost constraints comprise server-free function budget and virtual machine budget; integrating the hyper-heuristic framework into a reinforcement learning algorithm; an improved reinforcement learning algorithm is adopted to solve the cloud workflow scheduling problem, optimal execution resources are selected, and an optimal scheduling scheme is obtained; and performing real-time scheduling according to the optimal scheduling scheme, and introducing a deviation feedback mechanism to monitor an execution error in real time. According to the method, the workflow scheduling problem is decomposed into a closed-loop optimization process of state perception and action decision, dynamic environment perception, multi-target tradeoff and online strategy optimization are deeply fused, the global search function is achieved, local optimization can be achieved, the algorithm complexity is low, and robustness is high.

The invention belongs to the technical field of operator generation of deep learning accelerators, and particularly relates to an operator automatic generation method and device for a deep learning accelerator. The method comprises the steps that an adjustable program is generated for an operator generation task of target hardware through a large language model according to reference implementation, the adjustable program comprises labeled decision parameters and a candidate configuration set thereof, and the decision parameters are related to architecture characteristics and execution performance of the target hardware; the candidate configuration set is a discrete value set or a continuous value interval of the decision parameter based on hardware constraint; constructing a solution space based on the mark, and searching an optimal instance in the solution space through a closed-loop optimization process; and continuously iterating and optimizing the operator based on the genetic algorithm.

The invention discloses a manufacturing task autonomous negotiation and execution method based on a large language model agent, and the method comprises the steps: constructing a production scheduling agent, an equipment management agent, a material distribution agent and a quality control agent, analyzing a natural language task instruction through the production scheduling agent, and decomposing the natural language task instruction into subtasks; each agent calculates a utility value based on the load rate, the resource matching degree, the estimated completion time and the historical success rate, and performs structured negotiation to achieve a task allocation consensus; a prediction-check-rollback architecture is adopted to generate an action instruction sequence, the sequence is compiled into a time Petri network transition sequence, and reachability verification is carried out based on hard security constraints; production environment data is collected in real time to trigger anomaly detection and re-negotiation, and a formalized security verification and causal anti-factual reasoning parameter updating mechanism is introduced. According to the method, unstructured instruction understanding, autonomous task planning, multi-agent collaborative decision and closed-loop optimization are realized, and the problems of real-time performance, safety and interpretability of a large language model in manufacturing control are solved.

The invention relates to the technical field of medical informatization and intelligent scheduling, and particularly discloses a ward nursing process prediction and dynamic scheduling method and system.The ward nursing process prediction and dynamic scheduling method comprises the steps that firstly, a digital twinborn simulation environment synchronized with a physical ward in real time is established, and a nursing process is simulated through a multi-agent network; carrying out multi-time sequence deduction based on the environment, generating a prospective scheduling instruction set, and issuing and executing the prospective scheduling instruction set; calculating a prediction mismatch degree index by collecting actual execution data and comparing the actual execution data with a plurality of dimensions of an expected state; when the index exceeds a dynamic threshold value, a meta control layer is triggered to carry out self-adaptive adjustment on simulation environment parameters, and the self-adaptive adjustment comprises reconstruction of an agent behavior logic weight or switching of a scene deduction mode; and deducing and generating an updated instruction set again based on the adjusted environment to form closed-loop optimization.

The invention discloses a multi-agent system for generating a trestle design model under the drive of a large model, and relates to the technical field of BIM and AI crossing. The working process of the system comprises the steps that after a user inputs an instruction, the system automatically verifies parameters such as span combination and bridge floor total length, and the number of steel pipe piles is deduced through bridge pier rank calculation; the architect agent adopts a hierarchical layout algorithm and completes component coordinate derivation in combination with a default table; a programmer agent generates modeling codes according to a priority rule, and the problems of parameter missing, family library mismatching and the like are solved through an error self-correction strategy. The system integrates a long-term / short-term memory mechanism, realizes design parameter tracing and modeling process closed-loop optimization, and remarkably improves the design efficiency and modeling automation level of a complex trestle structure.

The invention relates to the technical field of software engineering and automatic testing, in particular to an E2E scene test optimization method and system based on target man-machine cooperation. According to the scheme, an optimization target is manually set, an initial test case and a business process are input, and a test resource library is constructed; the system automatically analyzes a service path, generates a use case dependency graph and a key path, and determines a core scene through manual verification; a candidate optimization strategy is generated by the system, and a target strategy is determined through manual screening according to resources and priorities; the system carries out simplification, merging and scheduling optimization on the use cases, and deploys and executes the use cases after manual sampling verification of coverage; the system collects execution data in real time and generates an evaluation report, and whether a use case is supplemented or a strategy is adjusted is determined after manual analysis; and finally updating the resource library and the strategy model, establishing a dynamic feedback mechanism, and triggering continuous optimization during system iteration. According to the method, the E2E test efficiency and the software delivery quality are remarkably improved through man-machine cooperation and closed-loop optimization.

The invention provides a lean production-oriented self-adaptive digital intelligent scheduling method and system, and relates to the technical field of data processing, and the method comprises the steps: taking an initial resource allocation scheme as the basis of task sequence optimization, constructing a time detection line on a time axis, detecting the overlapping condition of different tasks in an execution period, and recognizing resource conflict points; generating a preliminary scheduling plan under the condition of meeting the equipment capacity and man-hour constraint; based on multi-source data of a production site, evaluating the preliminary scheduling plan, updating scheduling parameters, and generating an optimized scheduling scheme; and converting the optimized scheduling scheme into an equipment control instruction, and transmitting the equipment control instruction to a control terminal of a production unit to complete closed-loop optimization of production resources. Reasonable adaptation and efficient scheduling of production resources can be achieved, the production efficiency is improved, and resource waste and conflicts are reduced.

The invention provides a hybrid level enhanced random algorithm based on double-loop optimization and a cross-platform implementation method thereof, and belongs to the technical field of modeling and parameter, and the method comprises the following steps: S1, initialization; s2, setting algorithm parameters; s3, executing internal circulation, and executing local search under a fixed threshold value; s4, judging constraint conditions; when there is no constraint condition, generating a candidate solution; s5, calculating a target function value; s6, judging updating of the matrix according to the target function value; s7, returning to an outer loop, and adjusting a threshold value Th according to the frequency for receiving the new design and the number of iterations; s8, after M iterations are completed, the acceptance rate and the improvement rate are calculated; s9, updating the threshold value Th; and S10, judging a condition for stopping iteration. According to the method, the problems that a general construction method is lacked, local optimum is easy to sink and cross-platform deployment is difficult to realize in a multi-factor mixing level test design are solved.

Disclosed are a neuromorphic chip compiling method and compiling system. The compiling method comprises: acquiring a pre-trained model and a data set of a neuromorphic application; generating an initial neuromorphic chip architecture and a network spiking input; performing an optimization and adjustment operation, performing simulation, and evaluating performance; and determining whether an optimization objective is satisfied, and performing loop optimization until an optimal neuromorphic chip architecture is obtained. In the method, by means of iterative optimization and effective evaluation, an optimal neuromorphic chip architecture is efficiently explored in a vast design space, thereby improving design quality and efficiency. The present application further provides a corresponding compiling system, comprising a generator, a simulator, and a compiler, thereby achieving automated compilation and deployment, and providing support for the design of large-scale multi-core neuromorphic chips.

The invention discloses an air inlet casing full-process simulation closed-loop optimization method, which relates to the technical field of mechanical manufacturing and processing, and comprises the following steps: simplifying an air inlet casing model; selecting a plurality of key procedures from the whole procedures; according to process parameters and boundary conditions of the key process, grid division is carried out on the simplified model, and a corresponding finite element simulation model is constructed; sequentially simulating the finite element simulation model according to a key process sequence, correcting a simulation result according to deformation data of actual processing, transmitting the corrected simulation result to a next key process for model simulation until simulation and correction of all key processes are completed, and constructing a full-process simulation prediction model; performing iterative optimization on multiple parameters in the whole-process simulation prediction model to obtain an optimal parameter combination; and correcting the simulation model according to the deviation between the actual processing data and the simulation result under the control of the optimal parameter combination. According to the method, high-precision prediction of various complex manufacturing processes can be realized, and the manufacturing precision is improved.

The invention discloses an FPGA test case automatic generation system and method based on a large language model. The system comprises a demand input module, a demand structuring module, a large language model case generation module, a case confirmation module, a verification execution module and a closed-loop optimization module. The large language model can automatically analyze the unstructured demand document and extract key information such as function points, input and output, boundary conditions and the like, and manual one-by-one interpretation is not needed. A big language model is guided through Prompt Engineering to dismantle complex requirements into test key points which cannot be subdivided, and it is ensured that all logic branches are covered by tests. The large language model can dynamically generate diversified test cases based on test key points, including a normal process, an abnormal process and an edge scene. And in combination with FPGA design specifications, historical error cases and other plug-in knowledge bases, the large language model can generate test cases closer to actual hardware behaviors.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a systolic array scheduling processing method, device, equipment and medium, and the method comprises the steps: obtaining a data processing model, and compiling the data processing model into an initial processing task based on hardware architecture features of a systolic array, acquiring to-be-processed data and analyzing data characteristics of the to-be-processed data, monitoring a real-time operation state of the systolic array processing device, inputting the real-time operation state, the data characteristics and an initial processing task into a scheduling decision module to generate a scheduling strategy, executing the scheduling strategy to complete data processing, and collecting performance data; and updating optimization parameters of the compiling module and strategy generation parameters of the scheduling decision module based on the performance data. According to the method, the initial processing task is generated through compiling, the scheduling strategy is dynamically generated in combination with the data features and the running state, parameter updating is achieved through performance data feedback, self-adaptive closed-loop optimization of compiling and scheduling is formed, and the calculation efficiency and the energy efficiency ratio are improved.

The embodiment of the invention provides a drive optimization method and system based on hardware state perception and a readable storage medium. The method comprises the steps that system data are collected, processed and stored, optimal adjustment instruction data are obtained according to standardized system data, verification processing is conducted through association mapping and execution adjustment, the program running state is monitored, and if the number of exceptions exceeds a preset threshold value, fault diagnosis is triggered, and a troubleshooting report is pushed. Performing interactive processing according to system standard parameters, updating bottom-layer codes and permission application logic and verification processes, updating mapping rules according to adjusted effect data, and optimizing a fault correlation analysis model; through hardware state perception and data acquisition, driving parameter dynamic matching and adjustment, fault intelligent diagnosis and report generation, cross-system version compatibility adaptation and data closed-loop optimization and rule iteration, balance between performance and resource consumption is realized, equipment operation stability and cruising ability are improved, and development and maintenance cost of a driving program is reduced.

The invention relates to the technical field of software testing, and discloses an LLM and SMT solver-based test case generation method, which is characterized in that a search type test is taken as a core drive, when coverage stagnation occurs, an execution path is subjected to semantic analysis by means of an LLM, a logic expression for describing a target path condition is generated, and in consideration of the limitation of the LLM in logical reasoning, a test case is generated; the system further introduces an SMT solver to perform verification and reasoning on a logic expression output by the LLM, finally feasible test input is generated, and the test efficiency and the coverage rate are improved. According to the method, the efficiency and precision of path coverage are improved, good adaptability and universality are achieved, the method is suitable for various languages and complex program structures, and the intelligent level and engineering practical value of automatic testing are effectively improved. The problem of insufficient accuracy of a large language model in complex path constraint reasoning is effectively relieved.

The invention discloses a gypsum dehydration control method and system based on multi-parameter collaborative optimization, belongs to the technical field of gypsum dehydration control, and solves the problem of how to improve gypsum dehydration efficiency under a dynamic working condition. Comprising the steps of collecting key operation parameters, a parallel operation mechanism model and a data driving model in the gypsum dehydration process in real time, constructing a weighted multi-objective optimization function based on constraint conditions, performing real-time optimization by adopting model prediction control, and performing global multi-objective optimization by adopting a periodic evolutionary algorithm. Generating a final control instruction based on the fusion strategy and issuing the instruction to an execution mechanism; by combining multivariable real-time prediction, closed-loop optimization control and a self-adaptive learning mechanism, intelligent management is carried out on the dehydration process of the wet desulphurization byproduct gypsum, accurate control over the water content of a filter cake, optimization of washing water and energy consumption and high adaptability of a control system to dynamic working conditions are achieved, and the dehydration efficiency of the wet desulphurization byproduct gypsum is improved. Therefore, on the premise of ensuring the quality of the gypsum, the dehydration efficiency is maximized and the resource consumption is minimized.

The invention provides a stability and bug testing method after updating of a game operation system, and belongs to the technical field of game stability and bug testing. S2, performance anomaly detection and collapse prediction; s3, Bug text classification is carried out; s4, generating an intelligent test case; s5, carrying out performance optimization related detection; s6, defect and anomaly analysis; and S7, performing test execution and closed-loop optimization. According to the invention, the stability and the accuracy and effect of bug testing are improved.

The invention discloses an interactive front-end generation method and system based on API specifications and multiple intelligent experience certificates, and belongs to the field of front-end software development automation. A double-stage intelligent generation framework is adopted, natural language query and back-end API specifications of a user are received, an optimal retrieval stage and an iteration generation stage are executed in sequence, and a code version with the highest comprehensive score is selected from iteration history to serve as a generated interactive front-end prototype code. Wherein in the retrieval stage, related UI components, icons and general layout structures are accurately retrieved to provide a high-quality context; a closed-loop optimization process dominated by an encoder and coordinated by multiple agents is started in iteration generation, the code quality is verified in a multi-dimensional mode, and the encoder agents are driven to conduct multi-round iteration correction on the code according to the verification result. And finally, the prototype code preferentially output can perform real data interaction with a back-end API, and can be used as a Demo or a starting point for rapid development, so that the front-end development efficiency is remarkably improved.

The invention relates to the technical field of water conservancy project automation and industrial control, in particular to a water conservancy equipment collaborative operation system based on an industrial control system, which comprises a programmable input mapping and logic matching unit, a data processing unit and a data processing unit, the system carries out binary bit mode matching on the operation result and the feature data block in the memory; the logical operation and addressing unit is used for calculating control lag time and generating a structured digital control sequence data stack; the sequential control and instruction addressing unit is used for generating a cooperative control instruction set containing sequential logic; the output mapping and checking unit is used for reading real-time feedback, and if deviation exists, an interruption compensation program is triggered to correct an action curve of an executing mechanism; the storage rewriting and loop optimization unit is used for correcting transmission delay parameters of the model and realizing closed-loop optimization; according to the method, the problem of model distortion caused by physical property drift is solved, and stable full-life-cycle operation is ensured.

The invention discloses a development management system and method for an API interface, and relates to the technical field of software engineering. The system comprises: an intention understanding and design engine, which analyzes a multi-modal service demand into a structured API development task; the agents collaboratively develop a network, and the multiple AI agents are dispatched to generate API specifications, codes, test cases and deployment configurations in parallel; the API ecological dynamic map construction module is used for automatically constructing and updating a dynamic map of dependency and data flow direction between APIs based on a generated product; a continuous learning and evolution engine analyzes the atlas and operation data to identify optimization points and generate or execute evolution policies. All the modules form a closed loop, and output of the evolution engine is fed back to the intention understanding engine to optimize subsequent design. The corresponding method comprises the steps of demand analysis, intelligent collaboration, atlas construction and closed-loop optimization. According to the method, automatic development, intelligent collaboration and continuous architecture optimization of the API are realized, and the development efficiency and the system maintainability are improved.

The invention provides a controller program generation method based on semantic planning and quality optimization, and belongs to the technical field of automation. A multi-stage PLC program generation and optimization method oriented to natural language task description is adopted, and semantic retrieval enhanced task planning, structured code synthesis and a quality closed-loop optimization mechanism based on feedback are fused. A natural language task is automatically converted into a verifiable plan with logic constraints by constructing a semantic index, reasoning scheduling and an intermediate representation structure; through module clustering and multi-round prompt guiding type logic reconstruction, generation of a high-quality PLC program meeting the IEC 61131-3 standard is realized. And static analysis and test feedback are combined, and the code quality is evaluated and enhanced by adopting iterative optimization and a scoring mechanism, so that the problems of intelligence, structuring and verification of the whole process from natural language to PLC program are systematically solved, and the development efficiency, the program robustness and the feasibility of industrial deployment are remarkably improved.

The invention relates to the technical field of intelligent power grids, and discloses a power transformation maintenance plan adaptive closed-loop optimization management method based on multiple elements, which comprises the following steps: constructing a multi-element database of power transformation equipment; based on the multi-element database, the maintenance priority of each device is calculated through a weight quantification model, and a maintenance priority sequence is generated; a mixed parthenogenesis optimization algorithm is adopted, the maintenance priority sequence is used as input, and an optimized maintenance plan is generated with the purpose of minimizing the comprehensive cost; and on the basis of an improved partially observable Markov decision process framework, modeling is performed on the execution process of the optimized maintenance plan, and a reinforcement learning mechanism is utilized to dynamically adjust and optimize a subsequent maintenance plan according to a feedback reward after the maintenance plan is executed, so that closed-loop iteration is formed. According to the method, adaptive dynamic optimization and sustainable evolution of the substation maintenance plan are realized, and the intelligent level, the safety reliability and the resource utilization efficiency of power grid operation and maintenance are remarkably improved.

The invention provides a macro-cell layout evaluation optimization method based on a large language model, and belongs to the field of electronic design automation. According to the method, a large language model is introduced to assist layout generation, a current initial layout is improved in combination with connection weight analysis, on the other hand, an improved result is judged and continuously iterated according to natural language feedback of an experience manual or an expert, and closed-loop optimization is formed; carrying out weight quantization on direct and indirect signal paths among macro units, constructing a logic association strength matrix, and realizing accurate identification and priority ranking of key time sequence paths; the availability score of each candidate layout scheme is calculated through a layout availability scoring module, dynamic measurement is carried out on surrounding standard unit resources, the resource distribution state is evaluated in real time, the position of a macro unit is actively adjusted, and resource congestion or waste is avoided; the method is deeply integrated with an open source EDA tool, the dependence on a commercial tool is reduced, the tool chain deployment and iteration period is shortened, and the transportability and usability of the method are guaranteed.

The invention discloses a MaaS platform construction method and system based on evaluation-driven closed-loop optimization, and the method comprises the steps: enabling a model square to serve as a unified entrance of a model, and providing rich model resources for a user to select; the computing platform provides model development, training and deployment for a user; evaluating the performance of the model; through feedback, close cooperation among the model square, the calculation platform and the model evaluation is realized, and an automatic optimization closed loop is formed. According to the closed-loop optimization mechanism based on evaluation driving, close cooperation among the model square, the calculation platform and the model evaluation module is achieved through feedback of the model evaluation module, an automatic optimization closed loop is formed, and the problem that an existing platform cannot achieve the automatic process from model selection, evaluation to retraining is solved.