208 results about "Decision fusion" patented technology

The invention relates to the technical field of enterprise management, in particular to an intelligent decision support system and method based on cognitive logic and scenarized semanteme, and the method comprises the steps: obtaining enterprise decision data through multi-source data monitoring, carrying out the preprocessing, and generating cross-modal enterprise scene cognitive information; analyzing cross-modal association among the data, fusing a cognitive logical reasoning rule and a semantic understanding model, and constructing scenarized semantic decision fusion features; training an enterprise decision-making quality evaluation model based on historical cases, performing quality perception on a current decision-making scene, and outputting decision-making quality information; and an execution effect is judged in combination with an expected target, an optimization mechanism is started if the target is deviated, candidate schemes are generated by using a historical knowledge base and a multi-target optimization algorithm, an optimal solution is screened, and intelligent adjustment of a decision scheme is realized. According to the method, multi-source heterogeneous data and cognitive logic are fused, semantic understanding, dynamic evaluation and adaptive optimization capabilities of a decision system are enhanced, and scientificity and real-time performance of enterprise decision are improved.

The invention discloses an intelligent management system and method for knowledge graph quality evaluation and self-repairing, belongs to the technical field of knowledge graphs, and aims to solve the problems that in traditional knowledge graph management, manual auditing efficiency is low, an effective automatic repairing means is lacked, and data complexity and real-time changes are difficult to deal with. The system firstly collects multi-source heterogeneous data in a target field, cleans the data through a deep learning noise recognition model, extracts entities and relationships by using a natural language processing technology, and adds metadata to convert the entities and relationships into graph structure data; then, a graph framework is defined based on the ontology, entity semantic alignment is achieved in combination with a graph neural network, and a knowledge graph is constructed by complementing implicit relations with the help of a pre-training language model. Then, the quality of the atlas is quantitatively evaluated through a four-layer quality evaluation system, meanwhile, a repair scheme is generated based on vulnerability feature extraction, knowledge base matching and decision fusion, and intelligent self-repair is achieved; the map can be monitored in real time and evaluated regularly, a repair strategy and a knowledge base are optimized through reinforcement learning, it is ensured that the map is kept accurate and time-efficient for a long time, and the practical value is improved.

The invention discloses a method for judging an RTK abnormal value in an automatic driving integrated navigation system, and relates to the technical field of automatic driving high-precision integrated navigation, and the method comprises the steps: collecting RTK observation data, inertial measurement unit data, a wheel speed pulse signal and LiDAR point cloud data, carrying out timestamp alignment and coordinate system unification, and generating a fusion data value; calculating a carrier-to-noise ratio weight signal quality index of the satellite through the fused data value, and generating a signal quality report; and combining the signal quality report, the current satellite geometric accuracy factor and the vehicle motion acceleration, calculating a residual threshold, constructing a coriolis force compensated double-integral prediction model by using inertial measurement unit data, and predicting the position of the vehicle at the current moment. According to the method, multi-dimensional features such as satellite signal quality, vehicle motion state and residual analysis are fused through multi-source decision, and the anomaly detection capability of complex scenes such as urban canyons is effectively improved.

The invention relates to the technical field of hydraulic valves, and discloses a digital hydraulic valve intelligent diagnosis system which comprises a data acquisition module, a data preprocessing and label generation module, a feature fusion module, a classification diagnosis and decision fusion module and a service life prediction and suggestion generation module. Data such as pressure, vibration, displacement, flow and pollution degree of the valve are comprehensively captured through deployment of a multi-source sensor array at key positions of the digital hydraulic valve and a self-adaptive acquisition strategy, early fault feature omission is avoided, then through preprocessing means such as soft-hard hybrid wavelet threshold denoising and multi-sensor time alignment, the data precision is effectively improved, and the accuracy of the data is improved. Then, through dynamic-static layered feature fusion and an attention weighting mechanism based on GRU, different feature advantages under steady-state and fault working conditions are fully combined, the fault feature distinction degree is greatly enhanced, and the problem that similar faults are likely to be confused is solved.

The invention relates to the field of electric power monitoring, in particular to an electric power monitoring system intrusion detection method and system based on flow analysis. The method comprises the following steps: collecting network traffic, analyzing and recombining to obtain structured session data; time sequence behavior features and function code distribution features are extracted to construct a multi-dimensional feature set; inputting the feature set into a compliance rule base and a behavior baseline model in parallel, and respectively outputting a rule matching result and an abnormal deviation degree score; generating a comprehensive threat index by adopting a weighted decision fusion strategy; and when the index exceeds a dynamic threshold value, intrusion is determined and an alarm is given. According to the invention, the problem of insufficient precision and adaptability caused by single feature dimension and isolated detection mechanism is solved.

The invention discloses a multi-modal data fusion and fault diagnosis method, and belongs to the field of transformer partial discharge fault diagnosis. According to the method, for the problems of false alarm and missing alarm caused by data isolation and lack of effective integration in partial discharge diagnosis of the transformer, acoustic, infrared and visible light multi-mode data are synchronously collected, pixel-level space alignment is carried out based on feature point matching, time sequence synchronization is achieved through hardware trigger signals, and the fault diagnosis accuracy is improved. Multi-level fusion diagnosis of a data layer, a feature layer and a decision-making layer is adopted, including channel superposition to form a fusion diagnosis image, voiceprint features, temperature rise features and arc light or corona features are extracted and input into a feature fusion model to obtain an associated feature vector, decision fusion is performed through a support vector machine classifier and a D-S evidence theory, and a decision-making result is obtained. And outputting a final diagnosis conclusion, thereby realizing accurate and reliable diagnosis of the partial discharge fault of the transformer.

The invention relates to the technical field of grinding machine on-line detection, and discloses a grinding machine working process on-line monitoring system, which comprises a data acquisition module for acquiring vibration, acoustic emission and environment signals and generating an original multi-dimensional signal vector; the data preprocessing module is used for preprocessing the original multi-dimensional signal vector to generate a purified signal set; the working condition inversion module is used for performing real-time inversion on a technological parameter estimation value of current grinding machining; the anomaly detection module is used for calculating and quantifying an anomaly score of the anomaly degree of the current working condition based on the purification signal set; the adaptive adjustment module is used for generating a dynamic threshold value according to the process parameters and comparing an abnormal score to judge an abnormal state; and the decision fusion module fuses the abnormal state and the environment humidity and generates a decision instruction. According to the method, the material hardness of the machined workpiece is estimated in real time, and the abnormal threshold value is dynamically adjusted according to the hardness estimation value, so that normal signal fluctuation and equipment faults caused by switching of normal working conditions are effectively distinguished, and false alarms caused by process changes are avoided.

The invention relates to the technical field of voice emotion recognition, in particular to a man-machine interaction voice sensing method and system based on a gradient intelligent calling subnet pool. The method comprises the steps of obtaining an emotion data set; constructing a man-machine interaction voice perception model based on a gradient intelligent dispatching sub-network pool; the system comprises an acoustic clue sensing purification module, a layered acoustic essential coding module, a gradient harmony subnet pool module, a task specific feature extraction module, a focus and confidence joint calibration module, a self-adaptive optimization strategy module and a real-time reasoning and decision fusion module. Carrying out emotion decision making by utilizing the constructed human-computer interaction voice perception model; and outputting a decision result. According to the invention, through the acoustic clue sensing purification module and the layered acoustic essential coding, the problems of emotional information distortion and identity feature confusion caused by real environmental noise are fundamentally solved.

The invention relates to the field of network public opinions, and discloses a public opinion information detection method based on a heterogeneous large model, which comprises the following steps: acquiring a public opinion text, performing sentence segmentation, denoising and word segmentation preprocessing, and inputting the processed text into a detection model to output a harmful information category and an early warning level. The detection model is composed of a first large language model and a second large language model, and has the structural characteristics of cross-architecture semantic alignment, hierarchical knowledge distillation, field attention enhancement, multi-channel decision fusion and the like. Wherein the cross-architecture semantic alignment realizes hidden space sharing through bidirectional projection; the hierarchical knowledge distillation dynamically distributes weights according to task contribution of each layer; introducing domain bias to enhance semantic focusing by domain-enhanced attention; and the output of the two models is adaptively integrated and classified through multi-channel decision fusion. According to the method, the accuracy, robustness and reasoning efficiency of public opinion harmful information detection are effectively improved.

The invention relates to a station building safety monitoring method based on multi-model decision and edge calculation optimization, and the method comprises the steps: obtaining station building image data, carrying out the processing of the data through employing a customized image enhancement technology, and constructing a sample set; designing a plurality of deep neural network models, performing mixed precision quantitative perception training on the models by using the sample set, and deploying the models at edge equipment; performing preliminary safety state detection on the power distribution room image which is acquired and enhanced in real time, and integrating preliminary detection results through a multi-model decision fusion mechanism; a cloud edge collaborative self-learning closed loop is established, conflicting, low-confidence or false detection samples of an edge end are transmitted back, a large model is used for auxiliary labeling and incremental training, a new model is issued after performance verification, and continuous iterative optimization is realized. According to the method, image enhancement, multi-model cooperation, edge calculation and an online learning mechanism are fused, the detection accuracy of the potential safety hazard of the station building in a complex environment and the self-adaptive capability of the system are improved, and a reliable technical scheme is provided for intelligent operation and maintenance of the station building.

The invention discloses a concrete vibration sufficiency judgment method and system based on multi-sensor fusion. The method comprises the steps that first information of a vibration rod in concrete and second information in the concrete are obtained; the first information is motion state information; the second information is hydration reaction state information and comprises temperature change in the concrete; fusing the first information and the second information to obtain a fusion result; vibration sufficiency judgment is conducted according to the fusion result, and a vibration sufficiency judgment result is obtained; according to vibration sufficiency judgment, weighted decision fusion is carried out on the first information and the second information according to a preset weight, and judgment is carried out in combination with a self-adaptive threshold value and a state machine model; and generating operation guidance information based on the global region or the local sub-region according to the vibration sufficiency judgment result. According to the method, intelligent and accurate judgment on the concrete vibration sufficiency is achieved, the vibration effect can be evaluated more comprehensively and objectively, under-vibration or over-vibration is effectively avoided, and the concrete engineering quality is improved.

The invention provides a multi-modal computer vision data fusion method, and relates to the field of computer vision data fusion. The method comprises the following steps: 1, firstly, carrying out data alignment, obtaining a new image representation through pixel-level fusion, and then carrying out sensor fusion to integrate data into a uniform format for subsequent analysis; 2, feature fusion is carried out, firstly, feature splicing is carried out to serve as input of a model, then an attention mechanism is used to pay attention to more important modal information, and finally joint embedding is carried out to compare and analyze data; and step 3, finally, decision fusion is carried out, classification results are weighted through a voting mechanism, and then weighted averaging is carried out on data to obtain a final result. By processing heterogeneity, missing data and noise among different modals, data fusion is performed among different modals in advance, the fusion effect is optimized through a more efficient algorithm by secondary data processing, and the fusion efficiency is improved.

The invention relates to the technical field of industrial intelligent control, and discloses a multi-model collaborative optimization industrial MES intelligent control system. The system comprises a data analysis module, a knowledge enhancement module, a collaborative decision engine module, a decision fusion module and a control execution module. And the data analysis module executes tree structure-based data link analysis on the MES multi-source heterogeneous data, and constructs a production data tree structure with a multi-layer father-son relationship. And a knowledge enhancement module injects a domain knowledge template into each node of the structure to form a knowledge-enhanced production data graph structure. And the collaborative decision engine distributes the graph structure to a plurality of heterogeneous artificial intelligence models for parallel analysis through the unified service gateway. And the decision fusion module performs fusion processing based on confidence weighting on the output of each model to generate a comprehensive control decision instruction, and finally the control execution module drives the MES to complete production adjustment, equipment optimization and material scheduling. According to the system, the accuracy, the real-time performance and the interpretability of industrial intelligent decision making are improved.

The invention belongs to the technical field of nuclear power station maintenance management, and particularly relates to a nuclear power maintenance decision-making system and method based on multi-Agent cooperation. The system comprises an input layer, a multi-Agent cooperation layer, a knowledge support layer, a decision processing layer, an output and interaction layer and a feedback learning layer. The input layer receives initial work order information including equipment basic information and fault description; the multi-Agent collaboration layer allocates sub-tasks to predefined six types of professional Agents according to work order information and completes information interaction; the knowledge support layer comprises a nuclear power professional knowledge base, a historical maintenance database and a rule and regulation library; the decision processing layer performs conflict detection, negotiation and decision fusion on Agent output; the output and interaction layer provides a decision result display and man-machine interaction interface; and the feedback learning layer realizes comprehensive improvement of nuclear power maintenance decision-making efficiency and safety. The method has the beneficial effects that a multi-professional Agent collaborative network technical means is adopted, and the technical effects of maintenance decision cross-professional information instant sharing and efficient collaboration are realized.

The invention discloses a springback compensation control system for a metal stamping part, and belongs to the technical field of metal plate forming. The system comprises a stress memory prediction module, a fractal compensation analysis module, a phase change compensation regulation and control module, a compensation decision fusion module and an execution feedback optimization module. The stress memory prediction module extracts stress evolution characteristics of the stamping process through a multi-scale memory network and a time decay attention mechanism; the fractal compensation analysis module adaptively adjusts the compensation grid density based on the fractal dimension to realize optimal configuration of computing resources; the phase change compensation regulation and control module predicts the phase change behavior of the material through the thermal-mechanical coupling constitutive model and calculates the compensation correction amount caused by phase change; the compensation decision fusion module intelligently fuses three compensation strategies by adopting Bayesian reasoning and a spatial modulation function; and the execution feedback optimization module realizes continuous optimization of the system through online learning. According to the method, the influence of three dimensions of stress history, geometric complexity and material phase change is comprehensively considered, accurate prediction and compensation of the springback behavior are achieved, and the forming precision and production efficiency of the metal stamping part are remarkably improved.

The invention discloses a charging pile rectifier open-circuit fault diagnosis method based on a fusion algorithm. Comprising the steps of collecting typical IGBT open-circuit fault signals of a rectifier, performing noise reduction on an auto-encoder, performing time domain feature extraction, performing frequency domain feature extraction, constructing a multi-modal diagnosis model, introducing a D-S evidence theory, and generating a final decision according to diagnosis results of a time sequence modal model, a frequency domain image modal model and a multi-modal feature fusion model. And the direct current charging pile rectifier open-circuit fault diagnosis method based on interpretable multi-modal feature fusion and decision fusion is obtained. The system comprises and / or can operate the direct current charging pile rectifier open-circuit fault diagnosis method based on interpretable multi-mode feature fusion and decision fusion. According to the invention, the open-circuit fault of the direct-current charging pile rectifier can be diagnosed timely and accurately.

The invention relates to a cognitive load monitoring method based on multi-source physiological information fusion. The cognitive load monitoring method comprises the following steps that 1, physiological information data collection is conducted through a multi-source data collection module; 2, performing data processing through a data preprocessing module to obtain preprocessed electroencephalogram, eye movement and electrocardio data; 3, performing data layer fusion, feature extraction and feature layer fusion on the preprocessed data to obtain a feature layer fusion matrix; and step 4, constructing a CNN-BiLSTM-Transform hybrid neural network model, carrying out decision fusion, and carrying out iterative optimization on the model through evaluation. According to the method, the core basis problem of multi-source physiological information data fusion is solved, and a high-quality data basis is provided for subsequent feature extraction and model training; the accuracy and environmental adaptability of state evaluation are greatly improved; and the sensitivity and the recognition precision of the model to the dynamic change of the cognitive load are effectively improved.

The invention provides an aging-suitable health lighting method and system, an intelligent lighting terminal and a storage medium, and the method comprises the steps: carrying out the rule matching of the motion features of a human body target in a monitoring space based on a predefined fall behavior discrimination rule, marking the human body target meeting the fall behavior discrimination rule as a candidate fall instance, and carrying out the recognition of the candidate fall instance. Generating a corresponding rule matching degree score; inputting the motion features corresponding to the candidate tumble instances into a pre-trained tumble identification classification model, and outputting a tumble confidence score; performing weighted decision fusion on the rule matching degree score and the fall confidence score to obtain a fall judgment comprehensive score; and dynamically selecting an aging-suitable illumination strategy from a preset strategy library for lamp illumination based on the fall judgment comprehensive score. According to the invention, a cooperative linkage mechanism between fall monitoring and illumination control is constructed, illumination intervention can be automatically triggered at the critical moment when the fall risk occurs, and safe and accurate illumination support is provided in time.

The invention discloses a large model reasoning acceleration method and system based on adaptive operator fusion, electronic equipment and a medium, and belongs to the technical field of artificial intelligence. In order to solve the problem of poor operator fusion in large model reasoning acceleration, the invention provides a large model reasoning acceleration method based on adaptive operator fusion, which comprises the following steps: constructing a fusion revenue regression model; performing static graph structure analysis on the target large model, and determining a fusion sub-graph set according to the fusion mode set; extracting topological characteristics of each fusion sub-graph in the fusion sub-graph set; obtaining a target hardware type and performance parameters in target hardware operation; through the fusion income regression model, the fusion income of each fusion sub-graph in the fusion sub-graph set under the target hardware type and the performance parameter is determined; and according to the fusion income, determining a fusion strategy of each fusion sub-graph, and generating a corresponding fusion kernel. According to different model structures, hardware platforms and runtime loads, a fusion strategy can be dynamically decided, the operator fusion effect is improved, and higher reasoning efficiency and resource utilization rate are achieved.

The invention provides an automatic short message auditing method and device based on two-channel model back-end decision fusion, and relates to the technical field of prevention and control management, the method inputs short messages into an autoregression judgment model and a large-scale pre-training model, respectively generates sensitivity scores, optimizes weight distribution through cross validation, and improves the accuracy of short message auditing. The accuracy of short message content auditing is effectively improved, and the phenomena of misjudgment and missed judgment are reduced; a double-channel model structure and multi-dimensional information fusion are utilized, and the advantages of different models are combined, so that the auditing system can keep efficient and stable performance under various input situations, and the adaptability of the system to different types of short messages is enhanced; through back-end decision fusion, weighting processing is carried out on two paths of output, and the advantages of respective models are combined, so that the accuracy and robustness of auditing are improved, and efficient and accurate safety management is ensured to be kept in large-scale short message auditing.

The invention relates to the technical field of substation automation of a power system, discloses a monitoring information table intelligent generation method and system for a substation and a medium, and solves the problems that in the prior art, compiling of a monitoring information table depends on manpower, the efficiency is low, and the accuracy is poor. The method comprises the following steps of: analyzing a system configuration description file of a transformer substation to obtain initial information of signals of the whole substation, simulating to generate a signal displacement message with a unique identifier, and sending the signal displacement message to a station level network; acquiring actual signal data fed back from a telecontrol device and a monitoring host in response to the signal displacement message; taking the unique identifier as a correlation factor, matching the initial information with the fed-back actual signal data, and constructing a multi-source signal correlation mapping table; performing consistency verification and decision fusion on the multi-source description of each signal entry in the association mapping table to generate a unified and accurate signal description; and automatically generating a standardized monitoring information table based on the unified description and the associated signal attributes.

The invention relates to the technical field of intelligent medical treatment, and discloses a critical patient enteral nutrition tolerance dynamic evaluation device, which comprises a multi-source data acquisition module, a causal topology construction module, a causal path tracking module, a risk quantitative evaluation module, an early warning traceability module and a decision fusion module, synchronously collecting enteral nutrition and key physiological parameters of the critical patient, and obtaining multi-dimensional time sequence data; a causal topology is constructed for the multi-dimensional time sequence data, and a time-varying causal inference map is obtained; taking a preset gastrointestinal intolerance core index as a root node, tracking the map topological structure, and obtaining a key causal path; quantitatively evaluating the key causal path risk to obtain a path risk measure; interpreting a path risk measure threshold, generating a causal early warning signal, and reversely tracing to determine a priority intervention target; fusing the early warning signal, the key path, the risk measure and the intervention target, and outputting a clinical decision basis; according to the method, the efficiency of dynamic evaluation of the enteral nutrition tolerance of the critical patient can be improved.

The invention particularly relates to a target classification and uncertainty evaluation method based on semantic association evidence fusion, and the method comprises the following steps: 1, constructing and training a target fusion classification neural network, and calculating a Dirichlet distribution concentration parameter of each modal input data representing an evidence quantity; step 2, associating the generated Dirichlet concentration parameter with the evidence quantity, and calculating single-mode uncertainty; step 3, constructing a semantic association matrix and performing discount correction, exploring potential association and confusion relationships among different categories, and completing multi-source evidence fusion through a Dempster combination rule; step 4, integrating the global uncertainty of the fused evidence and the local uncertainty of the single mode to obtain final uncertainty evaluation; according to the method, the modal information can be effectively fused to obtain high-precision fusion classification, the uncertainty of fusion classification can be quantitatively evaluated, a basis is provided for improving the safety and interpretability of intelligent classification decision, and the method is suitable for the multi-source sensor decision fusion field of automatic driving, medical diagnosis and the like.

The invention discloses a power distribution state monitoring method and system for a power distribution network. Distributed node data of the power distribution network are collected; time domain and frequency domain features of the distributed node data are extracted, and load flow increment track feature vectors are obtained and comprise the time domain feature, the frequency domain feature and the load flow direction feature; inputting the load flow increment trajectory feature vector into a convolutional neural network model based on dual-channel and decision fusion, outputting to obtain a disturbance classification label, and marking an abnormal monitoring node; constructing a power distribution network topological graph, and performing neighborhood analysis on the abnormal monitoring nodes to obtain a neighborhood node set; calculating the similarity between the power flow increment trajectory feature vector of the abnormal monitoring node and the power flow increment trajectory feature vector of each distributed node in the neighborhood node set to obtain a disturbance propagation path; and collecting disturbance propagation path data, calculating a feature weight of the disturbance propagation path data, and combining the disturbance propagation path data and the feature weight to calculate a power distribution network operation risk index and divide risk levels.

The invention relates to the field of artificial intelligence, and discloses a charging robot intelligent scheduling and control method based on visual perception. According to the method, a double-layer architecture integrating global planning and local decision making is constructed; a central scheduling system generates a global conflict-free space-time path; each robot constructs a local three-dimensional semantic map in real time through a visual sensor, and predicts a dynamic target trajectory by using a long and short term memory network fused with a social force model; and when a collision risk is detected, triggering a decentralized negotiation protocol, autonomously re-planning a local obstacle avoidance path according to the dynamic priority, and synchronizing the state to a central system. The system comprises a global path planning module, a visual perception module, a trajectory prediction module, a risk decision module, a conflict negotiation module, a local re-planning module and a state synchronization module. According to the method, millisecond-level local response and global collaborative optimization are realized, and the safety, the task continuity and the system expandability are remarkably improved.

The application provides an integrated circuit overlay error measurement method based on information fusion, which is characterized in that, for a given diffraction or scattering type overlay error measurement mark, model-based overlay error measurement (MBO), experience-based overlay error measurement (EBO), model / model hybrid measurement (MBO+MBO) and model / experience hybrid measurement (EBO+MBO) are used to extract overlay error values. Based on the calculation of electromagnetic field simulation means, the measurement distribution and its uncertainty of different methods are estimated, the prior knowledge is provided to eliminate abnormal measurement results, and the fusion decision basis is provided. Further, the different measurement extraction calculation results obtained by the actual measurement signal are fused by using a hierarchical information fusion method to obtain the final overlay error result. The method provided by the application can realize high-credibility extraction and calculation of overlay error, and is suitable for overlay error measurement under non-ideal measurement conditions.

This application provides an age-friendly health lighting method and system, an intelligent lighting terminal, and a storage medium. The method includes: matching the motion characteristics of human targets in a monitoring space based on predefined fall behavior discrimination rules; marking human targets that meet the fall behavior discrimination rules as candidate fall instances and generating corresponding rule matching scores; inputting the motion characteristics corresponding to the candidate fall instances into a pre-trained fall recognition classification model and outputting a fall confidence score; performing a weighted decision fusion of the rule matching score and the fall confidence score to obtain a comprehensive fall judgment score; and dynamically selecting age-friendly lighting strategies from a preset strategy library for lighting based on the comprehensive fall judgment score. This application constructs a collaborative linkage mechanism between fall monitoring and lighting control, which can automatically trigger lighting intervention at critical moments when fall risks occur, providing timely, safe, and accurate lighting support.

The invention discloses a modal missing-oriented discrete emotion recognition method based on feature reconstruction, and the method comprises the steps: obtaining a to-be-recognized video clip, carrying out the preprocessing of the video clip, judging the integrity of a modal, and generating a modal tag; jointly inputting the video clip and the modal label into a trained discrete emotion recognition model to obtain a recognition result of an emotion category; the model comprises a feature extraction and embedding module which is used for carrying out audio feature extraction and video feature extraction and feature embedding on a multi-mode video clip containing an audio mode and a video mode to obtain embedded features; the gating bottleneck fusion module is used for sequentially carrying out feature compression, feature expansion, feature enhancement and gating fusion operation on the embedded features to obtain fusion features; the classification reconstruction module is used for carrying out emotion prediction on the fusion features to obtain an emotion prediction probability and carrying out weighted decision fusion to obtain a prediction result of an emotion category; and meanwhile, when the audio mode or the video mode is missing, reconstructing corresponding mode embedding characteristics.

The application discloses a prostate cancer diagnosis model based on multi-parameter ultrasonic images and a training method thereof. The prostate cancer diagnosis model comprises a convolutional neural network for extracting image features, a feature reconstruction algorithm module, a long short-term memory neural network for extracting ultrasonic contrast semantic features, and a decision fusion classification network. The training method comprises establishing a training sample set, then inputting the training sample into the prostate cancer lesion diagnosis model for training until the error between the predicted value output by the prostate cancer lesion diagnosis model and the real value of the lesion target classification converges, and a qualified prostate cancer lesion diagnosis model is obtained. The application can extract the optimal features in different parameter ultrasonic images by virtue of the advantages of multiple feature extraction networks, and can obtain a prostate cancer diagnosis model capable of extracting rich features, recognizing accurate results and being robust by processing the time sequence features of ultrasonic contrast through the long short-term memory network.

This invention provides a method, apparatus, device, and storage medium for multi-branch parallel simulation and decision optimization based on a large language model. It constructs a dynamically evolving battlefield state space by receiving multi-source real-time situational data. Situational analysis identifies battlefield emergencies, which are then used as triggers to invoke the large language model to generate multiple differentiated decision branches, each containing macro-level objectives and micro-level strategies. For each decision branch, an independent parallel battlefield simulation instance is constructed and allocated computational and simulation resources, driving synchronous parallel simulations to obtain real-time simulation states. Multi-dimensional dynamic evaluation is performed to obtain a dynamic comprehensive score. Based on this, branches with scores continuously below a preset threshold are dynamically pruned, and their released computational and simulation resources are weighted and redistributed to surviving branches to improve their solution accuracy. Finally, high-scoring, high-quality branches are extracted, and the large language model is invoked to perform decision fusion on their key action sequences, outputting the optimal combat path and command sequence.