76 results about "Hybrid reasoning" patented technology

The invention relates to the technical field of information data analysis. The invention provides a public policy case analysis knowledge graph fusion reasoning method and system. The method comprises the following steps: generating standardized preprocessing data; performing extraction processing on the standardized pre-processed data to generate a structured triple set; processing the structured triple set to generate a multi-dimensional knowledge graph; constructing a hybrid inference engine, and processing the multi-dimensional knowledge graph to generate inference result data; performing incremental updating processing of nodes and relationships on the multi-dimensional knowledge graph, and performing parameter optimization processing on the hybrid inference engine to generate an updated knowledge graph and an optimized inference engine; and reasoning result data are processed, and a visual analysis result is output, so that the problems of limitation of a rule engine on causal reasoning, semantic fuzziness and logic illusion of a large language model in the policy field and insufficiency of a single-field knowledge graph on cross-field interaction influence revelation are solved.

The invention discloses an intelligent storage system supporting multi-source heterogeneous data fusion management, which relates to the technical field of multi-source heterogeneous data fusion, and comprises a data source adaptation module, an intelligent data source adapter is configured in a DataWorks data integration module, multi-source heterogeneous data is accessed through the intelligent data source adapter, and the data source adaptation module is connected with the DataWorks data integration module; extracting a semantic feature vector and a technical feature set; the neural symbol hybrid inference module inputs the semantic feature vector and the technical feature set into a neural symbol hybrid inference engine, calculates a similarity matrix between multi-source heterogeneous data fields through a BERT-based neural network, and imports the technical feature set into a field knowledge graph constructed by Neo4j for symbol logic verification to generate a unified metadata model; according to the method, a semantic similarity matrix calculated by a BERT-based neural network is combined with symbol logic verification of a Neo4j knowledge graph through a neural symbol hybrid inference engine, so that automatic semantic alignment and logic consistency verification of multi-source heterogeneous data are realized.

The invention relates to a factory equipment operation and maintenance decision-making system based on multi-modal data fusion and hybrid reasoning, which comprises a multi-modal data fusion framework and a hybrid reasoning model, and is characterized in that the multi-modal data fusion framework is used for integrating multi-modal data and dynamically weighting fusion features through a cross-modal attention mechanism; the hybrid reasoning model comprises a rule reasoning module, a probabilistic reasoning module and an AI enhancement module, the rule reasoning module generates a generative rule based on an expert knowledge base, the probabilistic reasoning module processes uncertainty data by adopting a dynamic Bayesian modeling equipment state transition probability, and the AI enhancement module introduces an artificial neural network to optimize rule reasoning parameters; and different factory affair scenes are adapted in combination with transfer learning. The method has the advantages that the method is suitable for power stations, power distribution rooms, fans and other scenes, and fault diagnosis, service life prediction and dynamic maintenance strategy generation are combined.

The invention relates to the field of personal intelligent security evaluation, and particularly discloses a chain reasoning hidden backdoor vulnerability detection method of a vision-language-action model, which comprises the following steps of: respectively injecting micro pixel disturbance and rare character marks into vision and language input; on the basis of model autoregression prediction characteristics, designing a hybrid reasoning chain fusing normal reasoning steps and abnormal backdoor branches; adopting prefix tuning to take the hybrid reasoning sequence as a pluggable prefix injection model; and generating the vulnerability sensitivity of the abnormal action instruction through the systematic verification process detection model. Compared with an existing method, the method has the advantages that a nondestructive testing mechanism based on prefix adjustment and optimization does not need to modify model parameters or depend on training data, and the safety and reproducibility of detection are guaranteed; a multi-mode triggering mechanism is constructed, and the hidden vulnerability of the model in a complex scene is effectively revealed; the abnormal branches and the normal process are fused in a chain mode, and the defense capability of the model for the concealment logic offset can be systematically evaluated.

The invention belongs to the technical field of intelligent question-answering systems, and particularly relates to an intelligent question-answering inference system based on a knowledge graph, which is characterized in that firstly, a knowledge graph construction module fuses multi-source data to generate a structured graph, and after a user inputs a natural language question, a question-answering analysis module completes intention classification and entity disambiguation and converts the question into structured query; an inference engine module fuses symbol rules and graph neural network inference through a hybrid inference sub-module, and a dynamic weight adjustment sub-module optimizes weights according to errors and attenuation factors to generate an inference result; the knowledge updating module incrementally updates the atlas in real time and detects conflicts, the interactive interface module visually presents a result, and the evaluation optimization module iteratively optimizes parameters in combination with offline evaluation and online feedback. The whole process is from user question asking to result output, accurate reasoning and continuous performance improvement are achieved, and multi-field question and answer requirements are met.

The invention discloses a search intention recognition method and device, equipment and a storage medium, and relates to the technical field of reinforcement learning, and the method comprises the following steps: generating training data based on an original corpus; performing supervision fine tuning processing on the lightweight language model based on the training data to obtain a fine tuning trained lightweight language model; performing strategy optimization on the lightweight language model subjected to fine tuning training through an improved GRPO reinforcement learning algorithm to obtain a lightweight language model subjected to reinforcement training; constructing a double-model hybrid architecture based on a short thinking chain model and a thinking chain-free model in the lightweight language model subjected to enhanced training; and outputting a final sequence of the related documents of the search terms of the user through the double-model hybrid architecture. According to the technical scheme of lightweight language model fine tuning, reinforcement learning optimization and double-model hybrid reasoning, the target search scene recognition accuracy is improved, the real-time response speed is high, and training time consumption and reasoning resource requirements are reduced.

The invention is suitable for the field of power station equipment management, and provides a power station equipment health management system based on multi-modal perception and hybrid reasoning, and the system comprises a data perception layer which is used for obtaining and preprocessing multi-modal perception data of power station target equipment, and obtaining multi-modal feature data; the hybrid reasoning layer is used for taking the multi-modal characteristic data as input of a pre-constructed hybrid reasoning model to obtain an equipment fault diagnosis result; the decision management layer is used for generating an equipment health management decision according to the diagnosis result and outputting the decision; the hybrid reasoning model comprises a rule reasoning sub-layer, a deep learning sub-layer and a case reasoning sub-layer, the deep learning sub-layer adopts a lightweight Transform model, and the lightweight Transform model comprises a cross-modal attention enhancement module and a fault feature intensified training module. In conclusion, the hybrid reasoning architecture is constructed, so that the fault diagnosis accuracy of the power station equipment is obviously improved, the false alarm rate is reduced, and the technical problems of few unconventional fault samples and difficulty in diagnosis of the power station are solved.

The invention relates to a computer system utilizing a computer model. A method, a computer storage medium, a program product and an apparatus for training an isolated routing model, and a hybrid inference system are disclosed. The method for training the isolation routing model comprises the following steps: acquiring a training task set; obtaining a first model passing rate and a second model passing rate; generating a soft label based on the first and second model passing rates; generating, by the isolated routing model, a prediction result indicating a probability that the training task is to be routed to the first or second major language model; determining a loss function based on the prediction result and the soft label; and adjusting parameters of the isolation routing model based on the loss function. The first and second model passing rates are determined by sampling a plurality of times a result of performing the training task on the first and second large language models, respectively. According to the method, the isolation routing model can be efficiently trained, so that tasks can be accurately routed among a plurality of large language models to obtain balance between performance and cost.

The invention discloses an Internet of Things alarm root cause analysis method based on gas Internet of Things construction, and belongs to the technical field of gas pipe network safety monitoring. The method aims at solving the problems that in the prior art, alarm root cause analysis is low in efficiency and poor in accuracy, and particularly composite alarms are difficult to process. The method comprises the steps that firstly, multi-dimensional data such as alarms, equipment attributes, pipe network topology and historical cases are comprehensively collected and subjected to standardization processing; then, a dual-drive hybrid reasoning strategy is adopted, rule reasoning based on a knowledge graph and case reasoning based on a machine learning model are operated in parallel, two paths of results are subjected to weighted fusion, and candidate root causes are generated; particularly, aiming at composite alarms generated in a short time, a complex alarm disassembling module is used for carrying out hierarchical analysis on time sequence and space dimensions, and core influence factors are positioned based on dynamic weight calculation. The most innovative part of the method is that a closed-loop feedback mechanism can be constructed according to a final verification result of field operation and maintenance personnel, and dynamic weights of influence factors are automatically updated, so that self-adaptive optimization of an analysis model is realized. According to the method, the accuracy and efficiency of alarm root cause positioning are remarkably improved, and self-learning and evolution of the system are realized.

The invention relates to the technical field of artificial intelligence-natural language processing and knowledge graph construction, in particular to a method for intelligently analyzing policy content by using a graph model. The invention discloses a method for intelligently analyzing policy contents by using a graph model. The method comprises the following steps of: increment acquisition; a format decoding step; an entity identification step; a relation extraction step; a body mapping step; a map writing step; a hybrid reasoning step; and a service output step. By means of the above step cooperation mechanism, automatic grabbing, accurate analysis, tense modeling and intelligent reasoning are achieved in a multi-format and easily-changed policy data scene, the data integrity and query timeliness are remarkably improved, and the problems of missing grabbing, missing extraction and version mixing in the prior art are completely solved.

The invention provides a vulnerability discovery method and system based on symbol-semantic hybrid reasoning, and belongs to the technical field of software security. The method comprises the steps that a target program is analyzed, and an event sequence is extracted; converting the event into a symbol with a time sequence label and confidence, and constructing a symbol dependency graph; reasoning based on the dependency graph, calling a large language model to generate a semantic reasoning action when the certainty is insufficient, and updating the state after symbol verification; new rules are extracted from the inference chain passing verification through reinforcement learning, and self-evolution is achieved; and outputting a vulnerability report containing the reasoning path and the confidence coefficient. The system correspondingly comprises a time sequence probability symbol module, a symbol-semantic bidirectional coupling reasoning module, a reinforcement learning self-evolution module and a report generation module. According to the method, the preciseness of symbol logic and the generalization ability of a semantic model are fused, the accuracy, interpretability and adaptive ability of vulnerability detection are effectively improved, and the method is suitable for security audit and code review of a complex software system.

The invention discloses a project budget execution management and control method and system based on financial big data, and relates to the technical field of financial big data analysis, and the method comprises the steps: collecting project feature data, budget execution records and market dynamic indexes, after normalization processing, inputting into a neural symbol hybrid inference engine, and generating a structured anomaly diagnosis report by using dual-channel parallel processing; and analyzing the structured anomaly diagnosis report into a state vector of a Markov decision process, dynamically generating an action candidate set through a strategy network based on an Actor-Critic framework, evaluating an action value by using a reward function, and generating a decision instruction. According to the method, high precision and interpretability of anomaly diagnosis are realized, the problem of model black box in financial decision making is solved, and misjudgment caused by dependence on correlation is avoided. A diagnosis-decision-verification-optimization closed loop is formed, the scientificity and reliability of budget regulation and control and the self-adaptive capability of the system are improved, and the secondary risk is effectively reduced.

The invention discloses a disease medical case deep derivation system based on medical case cross-domain fusion, and the system comprises a data collection and preprocessing layer which is used for carrying out the collection and preprocessing of multi-source medical case data, and obtaining initial data; the knowledge fusion and representation layer is used for converting the initial data into structured knowledge and generating a target medical case knowledge graph with a unified semantic basis; the intelligent derivation and reasoning layer is used for performing deep mining and intelligent reasoning through the target medical case knowledge graph; the application interface layer receives medical case information of a patient and inputs the medical case information to the intelligent derivation and reasoning layer; and deriving the medical case information based on the intelligent derivation and reasoning layer to obtain a derivation result. The data coverage capability is improved based on multi-source heterogeneous medical case deep fusion, the analysis depth is improved based on multi-level deep derivation and hybrid reasoning, the reasoning capability is improved based on combination of semantic reasoning, case reasoning and deep learning, knowledge discovery is achieved based on intelligent mining of deep diagnosis and treatment rules, and meanwhile all-around clinical decision support is facilitated.

This invention relates to a segmented hybrid inference method for large language models based on uncertainty-driven approaches. The method includes acquiring current text data and historical state features to estimate the uncertainty index of the current segment; minimizing a unified scheduling objective function based on the uncertainty index to obtain a target inference pattern; performing inference calculations based on the target inference pattern to generate information contribution values ​​corresponding to key-value pairs; calculating the corresponding dynamic merging control probabilities based on the uncertainty index and information contribution values, and performing weighted merging or pruning on the key-value pairs to be merged to obtain compressed key-value pairs; defining a deviation metric and limiting the deviation metric to not exceed a preset upper bound determined by the dynamic merging control probability set and the uncertainty index; triggering a rollback process when the deviation exceeds this limit; otherwise, feeding back the compressed key-value pair state to the next segment for iterative iteration until the inference of all segments is completed; thereby reducing memory usage and inference latency while ensuring accuracy.

The application provides a language model inference optimization method and device. The inference optimization method comprises: in response to obtaining at least one request information requesting to use a preset language model to infer input information, dividing the first request information into multiple pieces of request information according to a preset length; batch-merging the multiple pieces of request information and second request information, performing zero-redundancy full-inference and mixed-inference on the multiple pieces of request information and the second request information by using the language model, and obtaining multiple first inference results, wherein the second request information is at least one request information obtained at the same time as the first request information or obtained in a full-inference or mixed-inference process on the multiple pieces of request information; and performing incremental inference on the multiple first inference results by using the language model, and obtaining an inference result output by the language model when inferring the input information. Through the above method, the occupation of the video memory during language model inference is reduced, and the inference efficiency of the language model is improved.

The invention discloses a domain model-based power grid fault knowledge reasoning method and system. The method comprises the following steps of: constructing a power grid domain model; acquiring and preprocessing real-time data of a power grid; forward reasoning is carried out based on the logic of'cause guide fruit '; performing reverse reasoning based on the logic of'deducing cause from fruit '; performing hybrid reasoning by combining forward reasoning and backward reasoning, and improving the reasoning precision through bidirectional cross validation; introducing a reinforcement learning optimization reasoning path; generating a fault handling scheme, and feeding back and updating the domain model; the system comprises a model construction module, an acquisition and preprocessing module, a feature extraction module, a forward rule reasoning module and a reverse rule reasoning module. According to the invention, rapid and accurate diagnosis of power grid fault reasons is realized; a reasoning path is automatically optimized by introducing a reinforcement learning algorithm, so that the system can perform adaptive adjustment, the fault diagnosis and disposal efficiency is continuously improved, and the dependence on manual intervention is reduced.

The invention discloses a synchronous phase modifier fault diagnosis method based on a knowledge graph and a large language model, and the method comprises the steps: obtaining the operation data of a synchronous phase modifier, and carrying out the preprocessing of the operation data, so as to form a unified corpus and a standardized data set; based on the unified corpus, using a BERT model to encode the corpus to obtain context perception vector representation, and based on the context perception vector representation, performing sequence labeling and relation extraction, and outputting structured knowledge data; based on the structured knowledge data, constructing a synchronous phase modifier fault diagnosis knowledge graph; based on a pre-trained large language model, performing model training and optimization by using the standardized data set, understanding and analyzing a fault phenomenon described by a natural language, and outputting a semantic analysis result; and in combination with the output of the knowledge graph and the semantic analysis result, performing hybrid reasoning to obtain a fault diagnosis result. The method has the advantages of high diagnosis precision, high diagnosis efficiency and the like.

The invention discloses an industrial control system for automatic production of a circuit board, and belongs to the technical field of circuit board production, the system comprises a multi-modal data acquisition module, a self-adaptive control algorithm engine and a prediction maintenance subsystem, the multi-modal data acquisition module is used for acquiring multi-modal data in real time, and the self-adaptive control algorithm engine is used for processing the multi-modal data; comprise process parameters, equipment state parameters, environment interference parameters and equipment health parameters; the self-adaptive control algorithm engine is used for processing the multi-modal data through a hybrid reasoning model and generating a control instruction for adjusting the equipment state parameters; and the prediction maintenance subsystem is used for processing the equipment health parameters through the equipment health index model, predicting the residual life of the equipment and giving a maintenance suggestion. Compared with the prior art, each device can be adjusted more efficiently and accurately, so that the device is suitable for automatic production of circuit boards.

The present application discloses a data processing method, which relates to the technical field of efficient hybrid reasoning methods, devices and storage media for large language models based on routing technology. The efficient hybrid reasoning of large language models based on routing technology includes: if a user request is received, determining the similarity between the user request and each historical request in the cache library; if there is no similarity exceeding the similarity threshold, determining the scoring indicators of each candidate processing model based on the accuracy scoring model, the scoring indicators including accuracy, response time and interface cost; determining the comprehensive score of each candidate processing model according to the scoring indicators; selecting a target model from each candidate processing model according to the comprehensive score, and routing the user request to the target model. The present application achieves the technical effect of optimizing the selection of large language models to achieve efficient response to user requests.

A hybrid inference system for a coding assistant utilizes a routing model to predict whether output generated by a large language model for a given prompt would be accepted by a user of the coding assistant. The routing model routes the prompt when the routing model indicates that the output generated by the large language model is likely to be accepted. The routing model routes the prompt to a local model when the output generated by the large language model is not likely to be accepted. The routing model is trained on the historical output generated by the large language model for various prompts and the acceptance or rejection of the output by users of the coding assistant.

The present disclosure provides an inference optimization method and apparatus for a language model, an electronic device, and a storage medium. The inference optimization method comprises: in response to acquiring at least one piece of request information requesting inference of input information using a preset language model, dividing first request information into a plurality of segments of request information according to a preset length; using the language model to perform full inference and hybrid inference on the plurality of segments of request information and second request information, so as to obtain a plurality of first inference results, wherein the second request information is at least one piece of request information obtained simultaneously with the first request information or obtained during the process of performing full inference or hybrid inference on the plurality of segments of request information; and using the language model to perform incremental inference on the plurality of first inference results, so as to obtain an inference result output by the language model when performing inference on the input information. The method reduces video memory occupancy during language model inference and improves the inference efficiency of the language model.

The invention discloses a river-sea direct ship design optimization method and system based on CBR-RBR hybrid reasoning, and relates to the field of ship design, and the method comprises the steps: obtaining a design target parameter input by a user and an initial weight of an evaluation index; according to the design target parameters, case retrieval is carried out in a river-sea direct wide, flat and narrow ship type design knowledge base, and an initial design scheme is obtained; determining a search space of an optimization variable according to a main scale and a ship form coefficient in the initial design scheme; performing Pareto optimization on the evaluation indexes in the search space of the optimization variables by using a multi-objective evolutionary algorithm to obtain a non-dominated solution set; determining a score sequence according to the non-dominated solution set and the initial weight of the evaluation index; and selecting the design scheme of which the score sequence is within a set range as a user preference scheme. According to the method, the preliminary design scheme of the river-sea direct ship can be efficiently and intelligently generated and optimized.

The invention discloses a chat record crime element automatic labeling method based on natural language processing, and relates to the technical field of electronic data intelligent analysis, and the method comprises the following specific steps: sequentially carrying out data preprocessing to analyze multi-platform data and uniformly encode, carrying out semantic analysis to identify crime intentions and entities, and carrying out entity association reasoning on crime elements; and marking elements by using a structured label, and finally carrying out risk scoring and generating a judicial analysis report so as to form a complete crime element analysis process. According to the method, multi-language processing, dynamic weight and hybrid reasoning technologies are integrated, multi-platform data are compatible, the multi-language analysis problem is solved, the criminal entity recognition precision is improved, and the labeling efficiency and accuracy are greatly improved; meanwhile, a full-link judicial system is constructed, crime elements are standardized, electronic evidence export is supported, case mastering is assisted through risk scoring and multi-dimensional reports, a rule base is dynamically updated to adapt to new crimes, the working intensity of judicial personnel is reduced, and reliable support is provided for network crime attack.

The application discloses an industrial intelligent agent system and method fusing digital twin and a four-element core, belongs to the field of artificial intelligence, and comprises a digital twin perception layer, a four-element fusion core layer, a hybrid reasoning engine and an interpretable report generator.The digital twin perception layer is used for acquiring real-time state data of a physical entity; the four-element fusion core layer integrates OWL ontology modeling, graph database storage, large language model interaction and multi-physical field agent model, and generates a four-element collaborative knowledge processing result; the hybrid reasoning engine performs parallel rule reasoning, graph algorithm mining, large language model reasoning and agent model reasoning and fuses to generate a comprehensive reasoning conclusion; and the interpretable report generator generates a multi-modal report containing a reasoning path and physical evidence.The application realizes interpretable tracing of a decision-making process through four-element collaboration, realizes millisecond-level physical quantity prediction through an agent model, realizes continuous evolution of knowledge through a dynamic evolution module, and significantly improves the reliability, real-time performance and adaptability of an industrial intelligent system.

The invention provides a model reasoning method, device and system based on a storage and calculation integrated hybrid calculation architecture, and the method comprises the steps: carrying out the prediction of to-be-processed target domain data on a simulation storage and calculation integrated unit based on a deployed universal model, and determining a prediction result; under the condition that the prediction result does not meet the preset condition, retrieving the to-be-processed target domain data based on a deployed database on the digital storage and calculation integrated unit, and determining a retrieval result; and determining a reasoning result corresponding to the to-be-processed target domain data according to the prediction result and the retrieval result. According to the method, hybrid reasoning of the prediction result of the universal model and the retrieval result of the database can be realized, the reasoning precision of the universal model for personalized data of the user is improved, storage and calculation are integrated, and the reasoning speed and energy efficiency are improved by reducing the data carrying frequency.

The application provides a symbol graph-oriented influence maximization method and system, and the method comprises the following steps: acquiring symbol graph data, and performing representation learning on nodes by using a symbol-aware attention layer to generate symbol-aware node embedding; a seed set representation module based on a variational autoencoder is constructed, a conflict-aware regularization term is introduced to punish negative relationships in the seed set and encourage positive cooperation; a double-path diffusion model is used to simulate the propagation dynamics of positive and negative influences to obtain node net influence; the modeling capability of a complex teacher model is transferred to a lightweight student model through knowledge distillation; finally, a hybrid reasoning strategy is used to output a seed node set with high net influence and minimum internal conflict. The application can effectively process positive and negative edge heterogeneity information in the symbol graph, accurately simulate the inhibition effect in the influence propagation, generate a high-quality seed set, and has important application value in the fields of social network analysis, viral marketing and the like.

This invention discloses a fault diagnosis method and system for new energy sources integrated into a distribution network. The method includes: acquiring distribution network operation data; establishing an initial fault type set based on the distribution network operation data; constructing a multi-level fault knowledge graph including equipment, system, and environmental layers based on the initial fault type set; designing a hybrid reasoning mechanism combining rule-based reasoning and probabilistic reasoning based on the multi-level fault knowledge graph; using the hybrid reasoning mechanism for fault diagnosis to obtain fault diagnosis results; calculating the fault diagnosis coverage rate based on the fault diagnosis results; and optimizing the fault diagnosis process. This invention can continuously improve the coverage and accuracy of fault diagnosis, effectively addressing the challenges brought by a high proportion of new energy sources integrated into the distribution network. The core advantage of this method lies in its dynamic adaptability and multi-dimensional knowledge fusion, enabling the fault diagnosis system to continuously improve itself as the distribution network evolves and new fault types emerge.

The invention discloses a standardized data partitioning processing system supporting a multi-source data format, which relates to the field of data processing and comprises a preprocessing and format unification module, an intelligent pre-analysis and strategy selection module, a hybrid reasoning depth partitioning engine, a partitioning result optimization and verification module and a structured data output module. The method supports compatible processing of multi-source data formats; dynamic division is carried out based on a chapter structure of text content and a logic unit, semantic association content is ensured, and cross-modal semantic segmentation is avoided; the universality of the partitioning strategy is improved; intelligent partitioning of multi-source data is achieved, the requirement for manual intervention is remarkably reduced, the processing speed and efficiency are improved, and the method is particularly suitable for a large-scale document processing scene; downstream applications such as knowledge graph construction and intelligent retrieval can be directly connected, and the data utilization efficiency is remarkably improved.

The invention discloses a hybrid reasoning BIM (Building Information Modeling) design achievement auditing method for complex standard specification provisions. The method comprises the following steps: S1, defining a grammar structure and an entity class in a standard specification ontology; s2, constructing a standard specification knowledge graph based on the natural language large model; s3, constructing a mapping table; s4, constructing a design result file and text description information; s5, generating an executable rule statement; and S6, performing hybrid reasoning based on a rule reasoning engine and a natural language large model cue word technology. The method comprises the following steps: disassembling a standard text by using a natural language large model cue word technology, and constructing a knowledge graph; the natural language standard specification text is converted into an executable rule, and compliance auditing of the BIM design result is supported; and a rule inference engine and a large language model are combined to perform hybrid inference, so that the BIM model auditing precision is improved.

The present invention provides a maneuvering target tracking method based on hybrid inference. This method, based on variational Bayesian and sequential Monte Carlo hybrid inference, first designs a proposed distribution based on initial variational parameters. This proposed distribution is sampled to generate samples. The generated samples are then used to calculate a surrogate evidence lower bound. Finally, the variational parameters are re-optimized using stochastic gradient descent, resulting in continuous iterative updating of the sampled particles and variational parameters. This method transforms nonlinear state estimation and maneuvering model parameter identification into an evidence lower bound optimization problem. Stochastic gradient descent is used to achieve joint optimization of target state estimation and model parameter identification, enabling simultaneous state estimation and model parameter identification. This effectively addresses nonlinear filtering issues and improves nonlinear maneuvering target tracking performance. The method is advantageous in that it can adapt to nonlinear dynamic systems with unknown model parameters.