167 results about "Graph inference" patented technology

The present invention relates to the related technical field of natural language inference. Disclosed is a large language model joint inference method based on a knowledge graph-enhanced chain-of-thought prompt, comprising: constructing a local knowledge subgraph; decomposing an original question text into S sub-question texts and concatenating the original question text and the S sub-question texts; inputting a concatenated question text into a graph inference model to obtain a weighted entity distribution; from the weighted entity distribution, extracting first G answer entities having the highest confidence level, using an inference transition matrix to trace an inference process of each answer entity, and generating an inference path from a question entity to the corresponding answer entity; and using the inference path to assist a large language model in predicting an answer to the original question text. By means of the method, the large language model can quickly and accurately find an answer to a question text.

The invention discloses a mobile terminal streetscape image real-time segmentation method based on a lightweight neural network, and relates to the technical field of image segmentation. The method comprises the following steps: firstly, carrying out 320 * 320 adjustment, Z-score standardization, adaptive histogram equalization and 3 * 3 Gaussian filtering preprocessing on an input streetscape image; then, an improved MobileNetV3 backbone network is used, and a five-scale feature map is output in combination with DropBlock regularization through eight feature extraction stages including depth separable convolution and an SE attention module; multi-scale features are fused through a U-shaped structure, and a fusion feature map is generated through up-sampling, element-by-element addition of dimension reduction low-layer features and an attention gating module; and during reasoning, outputting a segmentation mask by using a convolutional layer, Softmax and a conditional random field, and finally performing knowledge distillation, weight pruning, 8-bit quantization and TensorRT optimization. According to the invention, high-precision real-time street view segmentation is realized, the robustness is high, and the method is suitable for different devices and scenes.

The invention discloses a user tagging management and demand analysis system based on big data e-commerce, and the system comprises the following steps: a collection module which is used for collecting multi-modal behavior data, and generating a context-aware behavior representation vector; the label extraction module is used for constructing a content feature channel, a behavior feature channel and an emotion feature channel and extracting a multi-granularity label candidate set; the atlas construction module is used for constructing a label atlas according to the multi-granularity label candidate set; the intention reasoning module is used for inputting the label atlas and the current behavior representation of the user into an intention reasoning network and outputting a potential demand representation vector of the user; the clustering module is used for inputting the user potential demand representation vector into an improved grey wolf demand clustering algorithm for clustering analysis to generate a user demand group tag; and the structure output module is used for generating a user-label-demand ternary structure. According to the method, multi-channel modeling and the improved grey wolf algorithm are fused, and accurate identification of e-commerce user tags and demands is realized.

The invention discloses a zero sample anomaly detection method and system based on a triple perception learning enhanced visual language model, and relates to the field of computer vision, and the method comprises the steps: extracting global and local visual features from an input image; in the visual coding process, local features in a deep network are corrected through a spatial perception attention enhancement module, and fine-grained attribute text description is generated for abnormal visual features; performing deep semantic alignment on the attribute text description and the general text prompt through an attribute perception guide module; calculating the similarity between the enhanced visual features and the optimized text features, and generating a pixel-level abnormal segmentation map; in the inference stage, the segmented image is converted into a space attention weight through an anomaly perception reconstruction module, the space attention weight is fed back to a visual encoder to generate final global feature representation, and an anomaly score is calculated. According to the method, under the condition that a target domain training sample is not needed, the anomaly detection and positioning accuracy and generalization ability of the model under the scenes of industrial defect detection and the like are remarkably improved.

The invention discloses a cross-institution financial data federated learning modeling system and a privacy compliance verification method, and relates to the technical field of financial data processing, the system comprises a federated aggregation module used for a server to align financial time sequence characteristics of each institution through a timestamp hash alignment strategy, and a privacy verification module used for verifying the privacy compliance of each institution; aggregating the aligned financial time sequence characteristics by using a federated time convolutional network, and generating time sequence model parameters for capturing cross-mechanism global time sequence dependence; the federation reasoning module is used for generating a minimum spanning tree representation of a global enterprise association relationship based on a security multi-party computing strategy by the server side on the basis of a local sub-graph of the enterprise association relationship constructed by each institution client side; the server performs cross-mechanism graph reasoning by using a federated graph attention network and an aggregation graph embedding vector and combining minimum spanning tree representation, and constructs graph model parameters for identifying enterprise associated risks; and the model joint training module is used for the server to fuse the time sequence model parameters and the graph model parameters so as to determine global model parameters.

The invention provides a control system and method applied to a smart home terminal and a terminal, is applied to the technical field of smart home Internet of Things, and has the advantages that a user behavior prediction model driven by machine learning and a dynamic rule generation engine are introduced, so that the system can break through a traditional three-level linkage mechanical execution framework; an intelligent control closed loop with environment perception, intention reasoning and autonomous decision-making capabilities is constructed, multi-dimensional data fusion analysis from equipment state parameters and user historical operation tracks to real-time environment variables is realized, and a composite control strategy including equipment linkage sequence optimization, scene parameter dynamic calibration and abnormal operation self-correction is automatically generated. And in combination with a federated learning architecture, cross-user knowledge migration and group behavior pattern mining are realized, and the generalization ability and scene adaptability of a control strategy are continuously improved.

The invention discloses a Mangbar enhancement-based graph reasoning ultrasonic image few-sample target detection method and system, and relates to the field of medical ultrasonic image target detection, and the method comprises the following steps: a backbone network extracts the features of an ultrasonic image, and carries out the coding through the position information; inputting the encoded features into CC-Mama to capture long and short term context dependence; enhancing visual features through an attention mechanism, and extracting a global semantic pool; a spatial topological relation is obtained through graph reasoning, an interpretable sparse adjacent matrix is learned from visual features, and related object recognition connection is reserved; mapping semantic representation from a global semantic pool to each region, and learning a topological relation between anatomical structures by using a GCN (Graph Convolutional Network); the fused features are connected from long and short term dependency levels and topological relationship levels for further prediction. The method aims at improving the detection capability of few targets, fuzzy targets and low-contrast targets in a medical image Few-Shot target detection task and improving the detection precision.

The invention provides an intelligent agent-based automatic verification and rule matching system for insurance-waiting evaluation reports. The intelligent agent-based automatic verification and rule matching system comprises a report acquisition module which pulls to-be-verified and historical insurance-waiting reports; the auditing processing module calls an attention Bi-LSTM model which is subjected to reinforcement learning fine adjustment and is fused with knowledge graph embedding, and scores the report sentence by sentence; the rule management module is used for storing and analyzing an equal-guarantee 2.0 structured machine rule and carrying out secondary accurate matching on low-confidence sentences; the knowledge graph module is used for maintaining a multi-dimensional association graph of a standard family, a control measure and an evaluation item and providing graph reasoning; the agent decision-making module generates actions of passing, returning, manual rechecking or automatic correction according to the auditing state; the feedback learning module collects artificial recheck differences, an incremental training model, a strategy network and atlas vectors; and the storage module stores reports, models, rules, maps and logs in a multi-modal manner through a relation library, a map library and a vector library, and implements data governance and access control. According to the invention, the audit efficiency, accuracy and standard consistency of the evaluation report can be improved.

The invention provides a fault diagnosis method and system for operation and maintenance of an intelligent manufacturing system, and the method comprises the steps: firstly constructing an equipment association diagram which is composed of equipment component nodes and component association edges, then obtaining the operation monitoring data of each equipment component node, extracting node features, generating working condition time sequence features, and carrying out the operation monitoring of each equipment component node; meanwhile, edge features are extracted from operation monitoring data of nodes at the two ends of a component associated edge to generate cooperative change features, then a pre-trained graph reasoning model is called, fault propagation analysis is carried out on the equipment associated graph, the working condition time sequence features and the cooperative change features, and a fault propagation result containing an abnormal originating node identifier and a propagation path sequence is generated; and finally, based on the result, matching a historical fault case library, positioning a root cause component, generating an operation and maintenance scheme containing a maintenance operation sequence and parameter adjustment guidance, and pushing the operation and maintenance scheme to an equipment control system to trigger a fault repair operation, so that the fault can be accurately diagnosed, the root cause can be positioned, and the operation and maintenance scheme can be quickly generated. And the operation and maintenance efficiency and accuracy of the intelligent manufacturing system are improved.

The invention provides a high-power gearbox operation condition identification and risk early warning method, and belongs to the technical field of gearbox risk early warning based on deep learning. The method comprises the following steps: firstly, acquiring operation data of the high-power gear box under multiple working conditions based on multiple types of sensors, and preprocessing to form a time sequence data set; then, a multi-branch feature extraction model is designed, and depth features of specific working conditions such as tooth breakage, bearing faults, axial displacement and tooth backlash changes are extracted with priori knowledge as guidance; further constructing a gearbox operation condition recognition model, and realizing multi-label condition recognition and confidence evaluation through feature fusion based on an attention mechanism, Transform enhanced modeling and a graph reasoning network; finally, a multi-stage dynamic early warning module is designed, graded early warning is triggered by integrating the instantaneous state, the evolution trend and prognosis evaluation, and safe and stable operation of the high-power gearbox is guaranteed. According to the method, the recognition accuracy and generalization performance of complex working conditions and multi-fault concurrent situations are improved.

The application discloses a kind of based on triple perception learning enhanced visual language model's zero sample exception detection method and system, it is related to computer vision field, method includes: extracting global and local visual features from input image;Visual coding process is corrected local feature in deep network by spatial perception attention enhancement module, and fine-grained attribute text description is generated for abnormal visual feature;Through attribute perception guide module, attribute text description and general text prompt are deeply semantically aligned;The similarity of enhanced visual feature and optimized text feature is calculated, and pixel-level exception segmentation map is generated;Inference stage converts segmentation map into spatial attention weight by exception perception reconstruction module, and feedback is generated to visual encoder final global feature representation and calculates exception score.The method of the application significantly improves the accuracy and generalization ability of model in industrial defect detection and other scenarios without target domain training samples.

A method for recognizing maneuvering space targets based on a complex region graph transformer relates to the field of inverse synthetic aperture radar image processing. The present invention aims to solve the problem that rapid maneuvers of space targets can cause large posture changes in ISAR images, and since ISAR images usually exist in the form of complex values, general real domain networks can lead to the loss of key phase information in ISAR images. First, the present invention uses a contrastive learning module to train ISAR image block pairs to adapt to changes in the target ISAR image. Secondly, the PGT feature extraction module extracts local and global context features of the image blocks through a graph reasoning method and a Transformer framework to obtain a more effective representation of the target. Finally, the graph recognition module updates the features of the nodes and edges between nodes formed by feature embedding, and outputs the classification probability and classification results of the target.

The invention relates to a bridge degradation identification and maintenance decision support method and system based on a knowledge graph, belongs to the technical field of traffic infrastructure intelligent operation and maintenance and engineering information processing, and solves the defects of weak data fusion capability, opaque causal modeling and non-traceable decision support in the prior art. The method comprises the following steps: acquiring and preprocessing multi-source heterogeneous data to generate a standardized data set; semantic alignment is carried out, and candidate entities and relations are extracted; constructing a knowledge graph body and loading the knowledge graph body into a knowledge graph; a self-adaptive quantile threshold value is adopted to binarize the factor intensity and the disease severity sequence, and a smooth point mutual information and a Spearman correlation coefficient are fused to learn a causal weight; complementing the influence relation of factors on the bridge through graph reasoning and calculating the weight; and finally, calculating a bridge degradation risk score, and outputting a high-confidence influence path based on a shortest path algorithm. According to the method, multi-source data fusion and degradation causal chain explicit modeling are realized, and a scientific basis is provided for a bridge maintenance decision.

The embodiment of the invention discloses a test question output method and device based on multi-model collaboration, electronic equipment and a medium. A specific embodiment of the method comprises the steps of performing multi-source data fusion on a multi-source test question association data set, and then performing knowledge point association analysis to obtain a dynamic test question knowledge point graph; performing multi-hop graph reasoning query on the dynamic test question knowledge point graph to obtain a test question associated knowledge point set; performing model collaborative generation on the test question generation request information and the test question associated knowledge point set to generate an initial test question set; performing quality evaluation on the initial test question set to obtain a test question quality evaluation value set; dynamically adjusting the initial test question set to obtain an adjusted test question set and storing the adjusted test question set; and carrying out expansion adaptive test paper group rendering display on the adjusted test question set to obtain a test question test paper and printing the test question test paper. According to the embodiment, the test question quality can be improved, the test question generation efficiency is improved, the test question generation duration and display duration are shortened, the test question display quality is improved, and waste of display resources is reduced.

The invention provides a dynamic retrieval enhancement recommendation method based on graph reasoning, and aims to solve the problems that an existing recommendation system is difficult to capture dynamic interest changes of a user and the article recommendation effect is poor, and intelligent recommendation is realized by constructing a dynamic reasoning graph structure of user behaviors, memories and candidate articles in combination with a graph neural network. Specifically, a dynamic memory retrieval module is designed, and key modes are extracted from historical behaviors of a user to construct a memory library; the current state of the user, the retrieval memory and the candidate items are constructed into a heterogeneous reasoning graph, and dynamic classification and weight distribution are carried out on memory nodes through a graph attention network; meanwhile, a reinforcement learning mechanism is introduced, recommendation effect indexes serve as reward signals, and a graph reasoning strategy is optimized in an end-to-end mode; according to the method, the problems of user interest drifting and object exposure insufficiency in a recommendation system are effectively solved, the transparency of recommendation decisions is improved through an interpretable graph reasoning path, and the performance of the recommendation system and the user experience are improved.

The invention provides a pedestrian intention reasoning method fusing scene interaction features and hierarchical temporal modeling, relates to the technical field of automatic driving, and solves the technical problems of low pedestrian intention prediction accuracy and generalization ability in the prior art. The method comprises the following steps: acquiring man-vehicle distance data, self-vehicle speed data and scene picture data; semantic segmentation is carried out on the scene picture data, and the object attribute of each pixel is identified and identified to obtain a segmented image, bounding box coordinates and pedestrian posture key points; the bounding box coordinate is the position of the pedestrian in the segmented image; extracting scene time sequence features in the segmented image; based on the bounding box coordinates, the pedestrian posture key points, the pedestrian-vehicle distance data and the self-vehicle speed data, pedestrian motion intention features in space-time correlation are extracted; and performing feature fusion on the scene time sequence features and the time-space associated pedestrian motion intention features by adopting a hierarchical temporal strategy, and predicting a pedestrian crossing intention result. The method is used in the pedestrian intention reasoning process.

The invention relates to the technical field of crop remote sensing samples, and discloses a crop remote sensing image intelligent sample library construction method based on a knowledge graph. The method comprises the steps of collecting multi-source remote sensing image data, and generating an enhanced image data set through fusion and enhancement processing; and constructing a crop field knowledge graph integrating the crop growth model, the soil type and the climate condition. A knowledge graph is used as guidance, crop areas are recognized through semantic analysis, multi-scale features are extracted, and category attributes and associated contexts are deduced through a graph reasoning algorithm in combination with a relation path; and generating a sample label candidate set through deep learning of the graph attention network, screening and correcting the sample label candidate set by using an optimization algorithm constrained by the knowledge graph, and outputting high-quality sample labels. And integrating and storing labels and image data into a graph structure sample library, implementing a dynamic updating mechanism to adapt to new data, running a quality monitoring process driven by a knowledge graph, and adjusting a construction strategy according to result feedback to realize intelligent and efficient construction of the sample library.

The invention discloses a grassland desertification intelligent closed-loop management method and system based on multi-source remote sensing, and belongs to the technical field of ecological restoration and intelligent decision making. According to the method, technical breakthrough is realized by constructing an intelligent closed loop of perception, cognition, decision and optimization: firstly, multi-source data are fused to construct a dynamic ecological knowledge graph, and a desertification cause diagnosis report for quantifying the contribution rate of each driving factor is generated based on graph reasoning; calling a governance measure knowledge base to generate a scheme according to a diagnosis result, and optimizing a decision through digital twinborn rehearsal; finally, based on treatment effect feedback, the knowledge base is dynamically optimized through reinforcement learning; the corresponding system comprises a multi-source perception fusion module, an ecological cognitive diagnosis module, a decision rehearsal execution module and a self-learning optimization module. According to the method, the bottleneck of lack of causal diagnosis and system stiffness of a traditional method is overcome, the crossing from passive monitoring to active cognition and from static decision to dynamic self-learning is realized, and an innovative solution is provided for treatment of a degraded ecological system.

The invention belongs to the field of software engineering, and particularly relates to a deep learning compiler test case generation method based on automatic reasoning of a computational graph structure, which can be used for generating a deep learning compiler test case so as to improve the test efficiency of a deep learning compiler. The method is divided into two stages: in a computational graph structure learning stage, a text generation model is adopted to learn structure information of a computational graph; the relationship between operators in the calculation graph is the same as the relationship between words in a natural language, and each calculation path is similar to a sentence; and then all calculation paths in the calculation graph are extracted, so that the existing calculation graph structure information can be learned by utilizing an efficient text generation model, and the method is applied to generation of a new calculation graph. In the calculation graph reasoning generation stage, the calculation graph structure reasoning model obtained in the calculation graph structure learning stage is utilized, so that the next node is automatically reasoned according to the information of the current node, and a complex and diversified deep neural network model is formed.

The invention discloses an artificial intelligence model training and reasoning management system based on a knowledge graph, and the system comprises the following modules: a data analysis module which is used for collecting multi-source perception data and forming an entity set and an event set; the graph construction module is used for constructing a context knowledge graph based on the entity set and the event set; the context mapping module is used for establishing a reversible mapping relationship between the context knowledge graph and the original multi-source sensing data; the problem analysis module is used for carrying out semantic analysis on a problem input by a user and cutting the problem to obtain a context sub-graph; the reasoning chain generation module is used for generating a reasoning chain meeting time sequence consistency and causal consistency; the graph reasoning module is used for generating graph reasoning embedding based on the improved T-GNN model and outputting a question and answer result in combination with a question and answer generation model; and the feedback updating module is used for updating the context knowledge graph and the model parameters. According to the method, structured modeling and credibility closed-loop management of the question and answer reasoning process are realized.

The invention relates to the technical field of electronic manufacturing automation, in particular to an SMT equipment multi-machine cooperative control system based on the Internet of Things, which comprises a data acquisition and digital twinning module, a multi-agent reinforcement learning scheduling module, a knowledge graph diagnosis module and a central cooperative controller. The data acquisition and digital twinning module constructs a digital mirror image of a production line, a basis is provided for prediction and simulation, and the problem of data islands is solved; the multi-agent reinforcement learning scheduling module realizes self-adaptive scheduling of dynamic events of a production line based on a real-time state and a simulation environment provided by digital twinning, and solves the problem of production scheduling rigidity; the knowledge graph diagnosis module realizes deep root cause positioning of complex defects by utilizing domain knowledge and graph reasoning, and solves the problem of difficult diagnosis; and the central cooperative controller feeds a diagnosis conclusion back to the first two modules to form a complete intelligent closed loop of perception-analysis-decision-execution-learning.

The invention discloses a target-driven navigation method and device based on context awareness and imitation learning, and the method comprises the steps: recognizing an object instance of interest in an image based on a target detector DETR, and constructing an object graph; based on context perception graph reasoning, in the navigation process, dynamic context information such as images, actions and memories serves as guidance, object features are projected to hyperplanes of corresponding contexts by means of a TransH method at each time step, the object relation is dynamically learned, and an intelligent agent can better understand the complex environment. Based on visual representation of Transform, visual features and graph features are fused, and spatial semantic information of the environment is better captured. Based on generative adversarial imitation learning, a new dynamic reward function is designed, and an intelligent agent is helped to avoid a deadlock state in combination with environment rewards. Based on a standard asynchronous dominant actor-commentator algorithm, an effective navigation strategy is trained by using a new reward function, and the navigation success rate and efficiency of the intelligent agent in an unfamiliar environment are improved.

The invention discloses a sub-graph reasoning method fusing logic rule learning and attack semantic enhancement. The method comprises the steps that an input layer dynamically integrates knowledge graph topology and an AMIE rule base, an initial k-hop sub-graph is generated, and structured input is provided for attack chain mining; the sub-graph extraction module is used for executing double confidence filtering, screening high-value attack chains and applying dictionary filtering to enhance semantic reliability; the sub-graph coding module adopts an entity perception update layer and a relationship aggregation evolution layer of a dual-channel mechanism to collaboratively model the spatial-temporal characteristics of an attack chain; the relation reasoning optimization module is used for dynamically injecting high confidence rules and optimizing triple scores; and the training optimization module is used for implementing task perception negative sampling. According to the subgraph reasoning method fusing logic rule learning and attack semantic enhancement, based on inductive reasoning and semantic perception modeling, by taking subgraph modeling guided by a logic path as a core, attack chain rules with high confidence in a training graph are mined, and the understanding ability of a model structure is improved.

The invention provides a virtual power plant-power distribution network voltage dynamic cooperative control method and device, and the method comprises the steps: obtaining multi-source heterogeneous data to construct a mixed data set, and mapping the mixed data set to a quantum state space for feature extraction; fusing the quantum features, the satellite cloud picture and the voltage waveform data, and modeling by adopting quantum probability; generating a DERs output scene based on a quantum state superposition principle, searching and quantifying the voltage out-of-limit probability amplitude of each node, and generating a power distribution network risk thermodynamic diagram; constructing a hierarchical control decision; the hierarchical control decision is executed, then a causal logic chain of a voltage out-of-limit event is generated through an invariant graph reasoning network, the contribution degree of features to the control decision is quantified, and iterative optimization of model parameters is achieved in combination with a meta-learning mechanism. Based on the method, the invention further provides a voltage dynamic cooperative control device, self-adaptive response of distributed energy uncertainty is achieved, and a solution is provided for voltage stabilization of the power distribution network in a high-proportion new energy access scene.

The invention provides a knowledge graph construction method based on parameter calculation logic, and relates to the technical field of graph construction, and the method comprises the following steps: S1, basic data set creation: collecting and sorting domain knowledge data, carrying out systematic classification and organization on the domain knowledge data, and creating corresponding basic data sets for different categories; according to the knowledge graph construction method based on the parameter calculation logic, the knowledge graph with dynamic calculation and reasoning capabilities is realized through parameter calculation logic expression and parameter, formula and index model construction, and the limitation of a traditional knowledge graph in processing a mathematical calculation relation is solved; effective technical support is provided for intelligent decision making in the vertical field, domain knowledge is disassembled into parameters, formulas and attributes based on the first principle by establishing domain knowledge basic data, the knowledge graph is constructed through the entity-relation, and accurate graph reasoning support can be provided for downstream large models and intelligent application.

The invention discloses a vehicle information security identification method and device based on a behavior knowledge graph. The method comprises the following steps: determining the behavior knowledge graph; receiving a node exception event association graph from the vehicle end, wherein the node exception event association graph is determined by the vehicle end through an entity associated with the exception index; adding a suspicious edge connection relationship in the abnormal event association graph as a preset low-confidence abnormal relationship into the behavior knowledge graph to obtain an updated behavior knowledge graph; and based on the updated behavior knowledge graph, analyzing the connectivity of the relationship between the nodes by adopting a graph reasoning algorithm and a preset recognition rule, and recognizing a target attack chain matched with or similar to the known attack mode. Therefore, on the basis of preset recognition rule analysis and a graph reasoning algorithm, intelligent reasoning and automatic judgment are carried out on complex attacks matched with or similar to the known attack mode instead of simple threshold comparison, and the complex attacks implemented across different entities or functional subsystems can be effectively recognized.

The invention relates to an accompanying robot decision-making method and device based on a cross-modal knowledge graph, equipment and a medium. The method comprises the following steps: acquiring original multi-modal sensor data, and performing time alignment and perception feature extraction on the original multi-modal sensor data to obtain a perception unit set; fusing the perception unit set to obtain a time sequence semantic vector and a semantic candidate entity; updating the cross-modal knowledge graph, and performing conditional aggregation and read extraction on the updated cross-modal knowledge graph by adopting a graph neural network to obtain an emotional state vector with probability distribution, an environment semantic abstract and a trigger factor set; and in combination with the long-term user portrait and the preset task constraint, graph conditional decision is carried out to generate an accompanying robot execution instruction. By the adoption of the method, real-time decision making and long-term self-adaption of an accompanying scene can be achieved through multi-modal time sequence fusion and a time hyperedge-based cross-modal knowledge graph in combination with conditional graph reasoning and a personalized strategy.

The invention discloses an industrial operation compliance monitoring method and system, and belongs to the technical field of intelligent manufacturing and industrial automation. The method comprises the following steps: acquiring industrial operation data of employees by using a multi-modal sensor; constructing a digital twinborn model in a layered manner, and mapping the industrial operation data to the digital twinborn model; comparing the industrial operation data with standard operation data by adopting a dynamic time warping algorithm so as to detect industrial operation deviation; and constructing a behavior intention reasoning network based on a Transform architecture, fusing operation time sequence characteristics, environment state parameters and historical behavior data, and distinguishing the industrial operation deviations through multi-task learning so as to identify malicious operations. According to the method, the problems that traditional manual inspection is low in efficiency and malicious operation is difficult to identify are solved, and intelligent closed-loop management of industrial operation compliance can be realized.

The invention provides a graph generation method and device based on a text graph reasoning model, equipment and a storage medium, relates to the technical field of artificial intelligence, in particular to the technical fields of computer vision, deep learning, large models and the like, and can be applied to scenes such as content generation based on artificial intelligence. The specific implementation scheme is as follows: acquiring a map generation request of a user; reasoning the map generation request by adopting a pre-trained text map reasoning model to obtain a retrieval instruction adopted when a retrieval enhancement generation tool is triggered; planning a graph generation step based on the retrieval instruction by adopting the text graph reasoning model; and generating a target picture based on the picture generation step.

The invention discloses an unmanned system multi-intention recognition method and system based on inverse reinforcement learning, and belongs to the technical field of artificial intelligence and autonomous decision making. According to the system, a deep multi-intention inverse reinforcement learning framework is provided for solving the problems of multi-modality, fuzziness and dynamics of unmanned system decision-making intentions in a complex dynamic environment. By designing a multi-scale intention encoder based on Transform, state-intention combined dynamic reasoning is realized; constructing an end-to-end intention perception reward network, and adaptively fusing multi-intention features through an attention mechanism; and proposing an expectation maximization optimization strategy to realize collaborative optimization of intention reasoning and strategy learning. The system can automatically separate and identify various potential intentions from expert demonstration of mixed intentions, and learn corresponding reward functions and decision strategies. According to the method, the intention recognition accuracy and the strategy reproduction capability in a multi-intention scene are remarkably improved, and an effective solution is provided for behavior understanding and autonomous decision making of an unmanned system in a complex environment.