139results about How to "Strong explainability" patented technology

The invention relates to a credit distribution method in multi-agent cooperation, and the method comprises the following steps: obtaining joint data of a multi-agent system, the joint data comprising observation information and execution actions of each agent; the joint data are input into a discriminator model, and the discriminator model carries out modeling on an interaction dependency relationship among multiple agents based on an interaction dependency relationship modeling module of an attention mechanism; generating an auxiliary reward signal corresponding to each agent through a discriminator model; generating a fused reward signal based on the auxiliary reward signal and a global reward signal from the environment; and training and updating the multi-agent strategy model by using the fused reward signal so as to optimize the cooperation strategy of the multi-agent system. By introducing a credit distribution mechanism based on a discriminator model, automatic decomposition and optimization generation of individual reward signals are realized, so that the learning efficiency and stability of a multi-agent system in a complex cooperative task are remarkably improved.

The invention discloses a multi-modal perception and agent collaborative urban operation event handling method, which belongs to the technical field of artificial intelligence, and adopts the technical scheme that multi-modal front-end equipment is deployed to collect data, a fusion confidence coefficient is generated based on multi-source data, and an event work order is generated when the fusion confidence coefficient reaches a threshold value; the method comprises the following steps: classifying work orders by utilizing an urban management knowledge graph, determining responsibility subjects and disposal guidance, and determining priorities by integrating risks and regions through a weight model; abstracting various resources as disposal resources, and realizing work order and resource matching assignment under the constraint of response time limit; when a complex event condition is met, disassembling into sub-work orders according to a knowledge graph and carrying out parallel processing; the processing resources upload processing feedback, automatic verification is carried out based on computer vision and multi-source evidence, closed loop is qualified, if not, processing or evidence supplementation is carried out, and a closed loop result is archived and used for continuously optimizing the model and scheduling. The beneficial effect of the invention is that the urban operation event handling method based on multi-modal perception and intelligent agent cooperation is provided.

The application provides a large model security management method, system, device and medium based on a core particle architecture, which can be applied to the technical field of semiconductor integrated circuits. The method comprises the following steps: setting a physically isolated special security core particle as a hardware trusted root in a heterogeneous integrated system, and connecting the special security core particle with at least one target core particle running a large model through a bus; deploying a lightweight security supervision model with a smaller parameter quantity than the large model in the special security core particle; collecting behavior characteristic data of the target core particle running the large model in real time; performing inference by using the lightweight security supervision model, calculating an abnormality metric between the current behavior of the target core particle and a preset normal behavior model; updating a dynamic trust state of the target core particle based on the abnormality metric; and performing a hardware-level security response operation by the special security core particle in response to the dynamic trust state reaching a predetermined threshold.

The application discloses a ransomware identification method and system based on multi-source data fusion, and relates to the field of ransomware detection. The method obtains file system event data, system resource running indexes, process behavior data and industrial control system behavior data, identifies file features, system running behavior situation features, process features and industrial scene features from the data, judges whether each feature meets the essential behavior features of ransomware encrypted files, and divides the features into strong indication features, medium indication features and basic situation features. Different mapping methods are used to determine sub-risk scores for different features, and finally a weighted scoring mechanism is used to obtain an overall urgency score, based on which a ransomware attack is determined, and then suspicious process operations are automatically triggered and terminated in parallel, and network connections are blocked to prevent the spread of ransomware attacks. The application can accurately identify known ransomware and unknown variants, and has good interpretability.

This application relates to the field of data processing technology, and in particular to a data fusion method and system based on neural pathways. The method includes the following steps: First, acquiring electroencephalogram (EEG) data and eye-tracking (EMT) data, aligning them using timestamps, and preprocessing them separately; then, extracting EEG features based on the preprocessed EEG data; extracting EMT features based on the preprocessed EMT data; next, fusion of features based on the EEG and EMT features by downsampling the data; finally, evaluating the synergy of the fused features. This application constructs a complete evaluation system by building a framework from a stimulus-visual paradigm to a signal acquisition platform and analysis methods related to neural structures. The time delay between the most relevant points in the time domain of eye-tracking and EEG features is used as an indicator of brain-eye synergy. The difference between these indicators has interpretability based on physiological structures, constituting a novel evaluation indicator for brain-eye synergy when assessing dynamic visual acuity.

The application discloses a new correlation image recognition and machine learning method based on fuzzy monotony, and belongs to the technical field of artificial intelligence of pattern recognition and machine learning; the method is defined as FMMCA, which evaluates local fuzzy monotone correlation by comparing row vectors and column vectors of an image matrix pair by pair, then the local correlations are weighted and accumulated, and finally the global fuzzy monotone correlation between images is obtained. The application directly uses fuzzy monotone correlation analysis to replace classical correlation analysis for multi-view research, so that the problems existing in classical correlation analysis do not exist, and the fuzzy monotone method feature does not need to be measured statically by distance, but can be measured dynamically by interval change, so that the influence of some noise is offset, the performance is improved, a new fuzzy monotone machine learning method is formed, the optimization of a focus loss function is not needed, the parameters are few, the robustness is good, and the computing power is small.

The embodiment of the invention provides a muscle relaxation depth prediction anesthesia system based on electromyographic signals, and relates to the technical field of medical monitoring technologies. The system comprises an electromyographic signal acquisition module used for acquiring electromyographic signals; the signal preprocessing module is used for preprocessing the electromyographic signals; the frequency spectrum state analysis module is used for performing windowing and frequency spectrum transformation on the preprocessed electromyographic signals to generate frequency spectrum characteristics of all time windows, and calculating frequency spectrum inertia reflecting dynamic changes of the frequency spectrum characteristics based on a preset frequency spectrum inertia model; wherein the spectrum features comprise a median frequency and a total power; and the trend prediction and display module is used for generating and displaying a muscle relaxation trend prediction result in combination with the current muscle relaxation depth and the frequency spectrum inertia. According to the invention, the method solves a problem of low reliability of muscle relaxation depth, and achieves the effect of improving the prediction precision.

The application belongs to the technical field of data center network and distributed computing system optimization, and discloses a physical perception dynamic topology reconstruction method for large model hybrid parallel training, comprising the following steps: collecting training phase signals and network topology physical state variables, and constructing comprehensive effective bandwidth for the current training phase signal task; establishing a quantitative mapping relationship between the network topology physical state variables and the end-to-end delay, obtaining a logical hop number sensitivity parameter and a bandwidth sensitivity parameter, constructing a topology state index based on the comprehensive effective bandwidth, the logical hop number sensitivity parameter and the bandwidth sensitivity parameter, combining the topology state index to perform forward-looking evaluation on candidate actions, performing network topology reconstruction, recording the measured delay and network state after network topology reconstruction, and updating the parameters. The application significantly reduces the end-to-end communication delay and suppresses the long tail delay under the premise of ensuring the hardware physical deployability and control stability, thereby improving the large model training efficiency and stability.

The invention provides a lithium battery residual life prediction method and system based on data space position analysis, and the method comprises the steps: building a battery residual life space prediction model based on a convex polytope mechanism according to the basic measurement physical quantity of a lithium battery and the health condition parameter of the battery; determining the shortest length of the online trend as a spatial trend constraint according to the battery health condition parameter of the lithium battery; extracting spatial trend characteristics based on the characteristic parameters of the online operation data of the lithium battery to be detected, and determining a spatial trend straight line by combining spatial trend constraints; and determining a residual service life prediction value corresponding to the space trend characteristics of the lithium battery to be detected by integrating the battery residual service life space prediction model. By adopting the scheme, the defects of complicated operation and insufficient prediction result precision in the prior art can be overcome, and a battery residual life prediction result with high physical significance and high interpretability can be efficiently obtained.

The application relates to the technical field of active sonar detection, in particular to a transmission waveform selection method for target wake field disturbance detection, which comprises the following steps: applying uniform engineering constraints to selected engineering realizable waveforms to construct a candidate transmission waveform set of a detection system; equivalent the influence of target wake field disturbance on sound propagation to a time delay perturbation process of multi-path propagation, and establishing a target wake field disturbance sensitivity model accordingly; using the target wake field disturbance sensitivity model, calculating the sensitivity indexes of each candidate transmission waveform in the candidate transmission waveform set and sorting, taking the candidate transmission waveform with the largest sensitivity index as the optimal transmission waveform most suitable for target wake field detection; instructing the transmission end of the detection system to transmit a sound wave signal based on the optimal transmission waveform. The method can improve the observability of weak propagation time-varying characteristics caused by wake disturbance, and make the selection of the transmission waveform have stronger explainability and scene adaptation ability.

The invention discloses an energy storage power station collaborative planning method and system based on uncertainty propagation analysis. In order to solve the problems that an existing energy storage planning method is indefinite in physical mechanism and insufficient in interpretability, the planning method comprises the steps that operation data of a power system are obtained, and an analysis alternating current probabilistic power flow model used for establishing an analysis relation between system voltage statistical characteristics and line power statistical characteristics is constructed based on the operation data; constructing a propagation matrix representing the uncertainty propagation characteristics of the system based on the analysis of the AC probabilistic power flow model; performing spectral analysis on the propagation matrix to identify one or more dominant propagation modes which play a dominant role in the uncertainty of the system; and based on the one or more dominant propagation modes, determining a planning scheme of at least one type of energy storage power station, the category of the planning scheme including the access position capacity of the energy storage power station. According to the method, energy storage configuration is guided by using an uncertainty physical propagation mechanism, and the risk control effect, the calculation efficiency and the engineering interpretability of a planning scheme are remarkably improved.

The invention belongs to the technical field of underground water numerical simulation, and provides an underground water pollution migration simulation method and device, electronic equipment and a storage medium, and the method comprises the steps: constructing a three-dimensional octree grid, a control equation and a definite solution condition of a hydrogeological structure of a target region; carrying out discrete solution on the control equation in the three-dimensional octree grid by adopting a finite volume method taking a node as a center; constructing a post-processing approximation function and a local posterior error indicator on octree leaf subunits of the three-dimensional octree grid; determining a unit set contributed by a global error by adopting a Drfler marking strategy, then performing splitting or merging operation on leaf units, and performing grid updating by adopting a high-order interpolation or volume weighted average method; and according to a preset simulation termination condition, repeatedly executing underground water pollution migration simulation and finally performing visual display. According to the method, the simulation accuracy of water pollution migration is improved in a posterior error estimation mode.

This invention discloses a robust fusion method for large models based on semantically aligned fuzzy clustering. The method includes: first, obtaining a sequence of probability distribution vectors from the outputs of multiple heterogeneous large-scale pre-trained models for the samples to be processed; then, introducing non-negative reliability weights to weight this sequence to obtain an aggregated probability matrix, and constructing a graph Laplacian matrix accordingly; second, approximating the aggregated probability matrix into a fuzzy membership matrix and a semantic prototype matrix, constructing a cost objective function by combining the graph Laplacian matrix, and iteratively updating each matrix and weight until convergence; finally, solving for the maximum value of the converged fuzzy membership matrix to obtain the sample prediction category result. This invention can effectively suppress the influence of inferior models in unsupervised environments, significantly improving the accuracy and robustness of fusion prediction.

The present application relates to the field of agricultural environmental information technology, and discloses a remote control method for nitrogen and phosphorus loss amount of paddy field facing environmental data. The method comprises the following steps: obtaining historical nitrogen and phosphorus loss data, real-time meteorological data and water and fertilizer operation information of the paddy field; constructing a mechanism constraint sub-model combining nitrogen migration and transformation dynamics and phosphorus adsorption and desorption balance; performing multi-source feature extraction on the meteorological data to generate a standardized environmental vector; inputting the light residual convolutional neural network to output an initial prediction value; combining the mechanism model to generate a corrected prediction value conforming to the conservation of mass through the Lagrange multiplier method; when the prediction value exceeds the threshold value, generating a drainage gate opening degree and a fertilizer plan adjustment instruction, and delivering the instruction to an edge terminal for execution through a narrowband Internet of Things.

The invention provides a false data injection attack identification method based on an adaptive residual weighted PINN. The method comprises the following steps: constructing a physical equation describing a system operation state; a self-adaptive residual weighting network is constructed, and the self-adaptive residual weighting network takes sensor measurement data with noise or attack as input and takes the reconstructed system state as output; training the adaptive residual weighting network by using a gradient descent algorithm; mapping sensor data into a system state meeting physical consistency in real time by using an adaptive residual weighting network; and according to a weight coefficient and physical residual information which are calculated in real time by the self-adaptive residual weighting network, identifying a false data injection attack through a dual-criterion mechanism.

The invention discloses a layered fine-grained image forgery detection method and a layered fine-grained image forgery detection system. And the authenticity of the input image is preliminarily judged through the authenticity judgment module. Constructing a multi-branch pixel-level counterfeit positioning network, and obtaining multi-scale feature representation from low resolution to high resolution through a feature extraction backbone network; a high-resolution forged mask is obtained through pixel-level positioning network prediction; and the decision network constructs a coarse-to-fine reasoning path according to the multi-level label system, sequentially completes forgery property judgment, forgery type judgment, method-level traceability and platform-level traceability, and outputs a conditional probability of a corresponding-level label as a traceability result. The method has a unified multi-level label system, the systematicness and the accuracy of detection are remarkably improved, the source and the position of counterfeiting are visually presented through mask positioning and hierarchical classification, and the interpretability is high.

The invention provides a marketing display method and system based on multiple agents. Customized display schemes can be provided for different customer types, the transmission efficiency of effective information in the marketing display process is improved, and the user experience is improved. In addition, the quality of products can be improved; in addition, the embodiment of the invention has the advantages of being high in interpretability and expandability, capable of supporting demonstration and questions and answers in professional fields and the like.

This invention provides a method for predicting the TCM functions of a single herbal drug and for drug screening. The method includes: treating each atom in the molecule of the herbal drug to be predicted as a node in a graph; constructing feature vectors for each node based on the physical and chemical properties of each atom; constructing a feature vector matrix based on the feature vectors of all atoms; treating the chemical bonds between atoms as edges in the graph; constructing an adjacency matrix based on the bonding types between atoms; constructing a graph structure matrix based on the feature vector matrix and the adjacency matrix; inputting the graph structure matrix into a graph convolutional neural network model to obtain a function score vector, the length of which is equal to the total number of preset TCM functions, and each score in the function score vector representing the probability that the herbal drug to be predicted belongs to a corresponding preset TCM function; comparing each score in the function score vector with a preset threshold, and determining the predicted TCM function value of the herbal drug to be predicted based on the comparison result. This invention improves the accuracy of function prediction.

This invention discloses a data modeling method for the thermal error mechanism of a worm gear grinding machine spindle considering the electromechanical-thermal coupling effect. The steps include: 1) establishing the correlation between the thermal expansion deformation of the spindle system and temperature variables; 2) establishing the thermal balance equation of the spindle system and deriving the correlation between the spindle thermal error and the heat absorption of the spindle structure; 3) constructing a mathematical expression for the heat generation of the spindle system with respect to the electromechanical-thermal variables; 4) constructing a mathematical expression for the heat dissipation of the spindle system with respect to the electromechanical-thermal variables; 5) establishing a theoretical model for the thermal error of the spindle system with respect to the electromechanical-thermal variables; 6) converting the solution of the spindle system thermal error model into an optimization problem, and solving it using the HPSO-GA optimization algorithm to obtain the thermal error of the spindle system. This invention can accurately predict the spindle thermal error while revealing the influence of the electromechanical-thermal coupling effect of the spindle system on the thermal error.

The application discloses a bearing fault diagnosis method and system based on a CS-SHAP model, bearing operation signals are collected through acceleration, temperature and acoustic emission multi-sensor, multi-dimensional fault features are extracted from time domain, frequency domain and time-frequency domain after wavelet transform or EMD algorithm denoising to complete preprocessing, feature data is divided into similar clusters by K-means clustering, SHAP values are calculated cluster by cluster and weighted according to cluster data density or diagnosis influence degree to obtain feature importance score, a SVM classifier based on RBF kernel function is trained with screened key features, model parameters are optimized through 5-fold cross validation, new collected data is preprocessed and input into the trained model to obtain diagnosis results, feature SHAP values are calculated and visualized by using the CS-SHAP model, and key factors of faults are analyzed. The application realizes accurate diagnosis of bearing faults and interpretability of diagnosis results, and provides technical support for intelligent operation and maintenance of industrial equipment.

The invention relates to a cognitive flashing system and method based on neural symbols, and the system comprises a multi-modal sensing module which is used for obtaining multi-modal data of a vehicle, and the multi-modal data comprises state data, software feature data and context data; the neural symbol fusion module is used for outputting a first flashing task schedule and a second flashing task schedule through a symbol inference engine and a neural network based on the multi-modal data; the flash task scheduling comprises flash target and flash resource allocation; the fusion module is used for fusing the first flashing task scheduling and the second flashing task scheduling through an attention mechanism and outputting a third flashing task scheduling; and the decision generation module is used for executing the flashing operation based on the third flashing task scheduling. By fusing symbolic reasoning and deep learning, the problems of stiffness and black box of a traditional flashing decision system are solved, and the adaptivity, safety and decision efficiency of vehicle software OTA flashing are remarkably improved.

This invention provides a lithium carbonate price prediction method based on multimodal data collaborative driving, comprising the following steps: multimodal data source identification and data acquisition; multimodal data classification and standardization; personalized noise filtering and outlier correction of multimodal data; extraction of trend and correlation features from structured data; extraction of semantic and visual / speech features from unstructured data; extraction of key information and feature quantification from semi-structured data; weighted collaborative fusion and feature optimization of multimodal features; construction and training of a two-branch collaborative prediction model; model prediction and preliminary verification of prediction results; optimization and dynamic correction of prediction results; interpretability analysis and visualization of prediction results; and method effectiveness verification and continuous improvement. This method, based on multimodal data collaborative driving, addresses the pain points of existing lithium carbonate price prediction technologies, forming a comprehensive, high-precision, and highly practical prediction solution.

The application discloses a high-frequency ignition online detection method based on an event camera, belongs to the technical field of industrial visual detection and event visual perception, and utilizes the event camera to asynchronously collect microsecond-level brightness change event streams, automatically locates multiple-point ignition ROIs through spatial event density combined with geometric constraint and periodic consistency constraint after preprocessing, constructs time event density curves in each ROI, accurately extracts ignition intervals and completes counting based on main peak detection and local half-peak boundary, and determines missing point abnormities in combination with a preset ignition timing, and determines multiple-point abnormities in combination with a single-ROI periodic peak value, a short-time simultaneous lighting quantity and interval overlap degree in multiple dimensions. The application has high detection accuracy, good real-time performance and strong anti-interference capability, and is suitable for online monitoring of high-frequency transient working conditions of an aero-engine ignition device and various multiple-point sequentially triggered ignition devices.

The application discloses a kind of based on knowledge graph's shot edge detection method, comprising the following steps: obtaining image data and detecting metadata, data preprocessing is carried out to obtain preprocessed image data;Current detection context is generated, and edge visibility evidence vector is obtained;Main strategy multi-path connection query is constructed;Leapfrog Triejoin algorithm is used to execute main strategy multi-path connection query, and main strategy configuration is output;Based on main strategy configuration and preprocessed image data, edge contour generation processing is executed, while consistency check is executed, and consistency check result and failure reason code are output;When consistency check result is not passed, back-off strategy multi-path connection query is constructed;Execute back-off strategy multi-path connection query, output back-off strategy configuration again Execute edge contour generation processing and consistency check, and output shot edge detection result.The application improves the detection precision of shot edge detection under complex defect conditions.

This invention relates to the fields of artificial intelligence and machine learning technology, and discloses a method and system for predicting the distribution of cadmium content in arable land soil based on machine learning. The method includes: acquiring multi-source environmental geographic data and measured cadmium content data; performing spatial alignment and scale unification processing; constructing a confounding factor identification model based on causal inference to screen true driving factors with causal relationships to cadmium content; inputting the true driving factors into an ensemble learning model to train the prediction model, and generating a high-resolution continuous spatial distribution map. The system comprises five modules: multi-source data acquisition, data preprocessing, causal feature screening, model training and prediction, and spatial mapping. This invention improves prediction accuracy, generalization ability, and interpretability by combining causal-guided feature screening with ensemble learning, providing scientific support for the safe use of arable land and pollution prevention and control.

This invention discloses a cross-modal data processing system for the safe operation of hydrogen refueling stations, including a raw data acquisition and processing module, a full-variable safety scanning module, a physical relationship coupling diagnosis module, an adaptive operating condition clustering module, a question-answer pair construction module, a hydrogen refueling station time-series command data acquisition module, and a fault type identification module. By constructing a three-layer semantic enhancement logic and model adaptation strategy, this invention effectively solves the technical problems in the prior art, such as the lack of supervision signals in the raw data of hydrogen refueling stations, the difficulty in identifying hidden faults under complex operating conditions, and the difficulty in adapting heterogeneous feature space models.

This invention discloses a method, medium, and equipment for constructing a multidimensional energy consumption quantitative analysis model for data centers. First, it uses multinomial regression to fit the relationship between the utilization rate of physical CPUs, memory, and disks and their power consumption. The power consumption of physical hosts, virtual hosts, containers, and computing tasks is considered as the sum of the power consumption generated by the CPU, memory, and disk. CPU utilization, memory utilization, and disk I / O throughput are used to represent the power consumption of physical hosts, virtual hosts, containers, and computing tasks. Finally, the energy consumption over a specified time period is calculated by integrating the power consumption over time. Compared with existing technologies, this invention studies the hierarchical relationship of data centers, comprehensively reflecting the overall energy consumption of data centers, and has the advantage of reflecting data center energy consumption from multiple levels and dimensions.

The invention provides a deep fake face image detection system and method, relates to the field of artificial intelligence safety, and aims to solve the problems of poor generalization ability, single detection dimension and insufficient robustness in the prior art. The method comprises the following steps: generating a three-dimensional geometric feature map according to a to-be-detected face image; performing feature extraction and cross-modal fusion on the original image and the three-dimensional geometric feature map to generate multi-modal fusion features; performing frequency domain and spatial domain analysis in parallel to generate frequency domain-spatial domain joint features; and inputting the two fusion features into a multi-engine decision module comprising a plurality of detection engines, and determining a final detection result according to the output of each engine through a decision fusion unit. According to the method, multi-dimensional information and multi-engine decision are combined, the accuracy, generalization ability and robustness of detection are effectively improved, and the method has interpretability.

This invention discloses an intelligent adaptive adjustment system and method for slurry supply in CMP equipment. The system learns the optimal position of the slurry supply arm for different process parameters from historical production data using a model. Measurements are taken of each wafer after grinding, and the results are fed back to the model to adjust the position of the slurry supply arm in real time. Furthermore, the thickness data at each sampling point on the wafer surface is spatially gridded and used as training data for the model, thus preserving the spatial information of the wafer surface in the samples and enhancing the interpretability of the model's predictions. Simultaneously, the use of two-dimensional feature encoding reduces the number of parameters, which is more advantageous for small sample datasets, especially in scenarios like wafer production where obtaining experimental datasets is costly. Ultimately, this invention achieves uniform distribution of the slurry on the grinding pad, effectively improving the surface morphology of the wafer after grinding.

The application discloses an intra-city traveler portrait system based on mobile signaling data, comprising a data preprocessing module, a travel chain extraction module, a travel mode recognition module, a daily travel mode recognition module, a travel path flow recognition module and a traveler portrait module. The system loads the dwell point data processed by the mobile signaling data supplier and combines the administrative division data for preprocessing, extracts the travel chain, recognizes the travel mode by using the Gaode map API and the log Gaussian mixture model, recognizes the daily travel mode by combining the convolution self-encoder and the K-means algorithm, counts the travel path flow, and finally summarizes the labels of the travel characteristics to construct the traveler portrait. The system optimizes the data processing flow, improves the accuracy of the travel mode recognition, enhances the interpretability of the travel mode recognition, and associates the macro traffic flow with the individual travel path, thereby providing a precise and efficient analysis tool for traffic demand management.