467 results about "Mixture model" patented technology

Provided are a method and an apparatus for constructing a road congestion prediction model, a device, a medium, and a product. A road traffic network is defined as a directed weighted graph. Historical dynamic traffic features of each road segment in the road traffic network are obtained as sample data, including recent dynamic traffic features and periodic dynamic traffic features. The sample data is input into a mixture of adaptive graph learners (MAGL) model for learning, and a probability prediction vector is output. The sample data is input into a trend expert model, and a trend distribution vector of a predicted probability of future traffic conditions is output. The periodic dynamic traffic features are fused to determine a periodicity prediction vector. An aggregated logit vector is obtained. An objective function is determined based on the aggregated logit vector. Congestion prediction training is performed to obtain a road congestion prediction model.

The invention discloses a method, a system and equipment for predicting the stability of integrated circuit test equipment and a medium, and belongs to the technical field of integrated circuit test. The method comprises the following steps: firstly, carrying out preprocessing and feature selection on historical data in an FT test stage, and adopting a Gaussian mixture model (GMM) to cluster and identify different operation condition clusters of a test machine; then, establishing a health state GMM reference for each working condition cluster, calculating a KL divergence value of a normal sample and the reference, and setting a dynamic anomaly detection threshold by 99.73% quantile of the KL divergence value; and finally, in real-time monitoring, calculating a KL divergence value of real-time data and a corresponding working condition cluster benchmark, and comparing the KL divergence value with a dynamic threshold value to realize accurate anomaly marking. The method effectively solves the problem of abnormal detection of the test data of the integrated circuit under complex and changeable working conditions, and improves the monitoring accuracy and working condition adaptability.

The invention discloses a transient electromagnetic and magnetic method data joint inversion method. The method comprises the following steps: acquiring transient electromagnetic observation data and magnetic method observation data; converting the transient electromagnetic observation data into transient electromagnetic moment data; based on the transient electromagnetic moment data and the magnetic method observation data, a joint inversion objective function is constructed, and the joint inversion objective function comprises a data fitting residual term and a model constraint term; performing parameter coupling on the model constraint term by using a multivariate Gaussian mixture model to obtain a coupled model constraint term; and combining the data fitting residual term and the coupled model constraint term into a final objective function, and minimizing the final objective function through an optimization algorithm to obtain an electrical parameter and a magnetic parameter of the underground medium. According to the method, transient electromagnetic and magnetic method data joint inversion is realized, the problem of low three-dimensional inversion calculation efficiency is solved by introducing transient electromagnetic moment, and electrical and magnetic parameters are coupled by utilizing rock physics constraint, so that the inversion result is more accurate, and the structure is more consistent.

The invention relates to the technical field of image processing, and discloses a pulmonary embolism focus segmentation method and system based on image recognition, and a storage medium. The method comprises the following steps: extracting a multi-level blood vessel topological structure of a CTPA image through blood vessel diameter gradient analysis; modeling blood vessel density distribution by using a Weibull mixed model to obtain embolism characteristic parameters; the pixel embolism probability is estimated through variational Bayesian reasoning, and a focus distribution diagram is generated; performing multi-scale feature fusion on the lesion probability graph to obtain a segmentation boundary; and obtaining a final embolism focus segmentation result based on the vascular connectivity constraint optimization boundary. The problems that blood vessel level differentiation processing cannot be achieved, and accurate probability modeling and anatomical constraint verification are lacked are solved. The accuracy of pulmonary embolism focus segmentation is improved.

The invention relates to the technical field of communication production management, and discloses a communication equipment production intelligent management system based on machine learning. The system comprises a production data acquisition module, a feature engineering construction module, a dynamic clustering analysis module, an anomaly detection engine module and a production decision optimization module. The production data acquisition module acquires multi-source sensor data in real time and converts the multi-source sensor data into a standardized sequence with a unified timestamp; the feature engineering module extracts a time domain statistical feature, a frequency domain energy feature and an equipment state association feature to generate a high-dimensional feature vector set; the dynamic clustering module adopts an incremental algorithm to divide clusters online; the anomaly detection module establishes a multi-level Gaussian mixture model based on a clustering label, and quantifies an anomaly probability through a mahalanobis distance; and the production decision module integrates the results to generate an equipment maintenance priority sequence and a production takt adjustment instruction. According to the system, intelligent monitoring and dynamic optimization of the whole production process of the communication equipment are realized, and the real-time change requirement of a complex production environment is met.

The invention relates to the field of computer vision, and discloses a defect detection method, system, medium and equipment based on adaptive dual-domain filtering and Gaussian mixture prior constraint, and the method comprises the steps: carrying out the multi-scale feature extraction of an ultrasonic C-scan image through a Vision Transform network after the ultrasonic C-scan image is preprocessed; respectively inputting the shallow fusion features and the deep fusion features into a frequency-space double-domain adaptive feature filtering module for filtering, inputting the filtered features into a Gaussian mixture modeling module Ada-GMM, and modeling normal feature distribution; carrying out Ada-GMM-Guided decoding, and carrying out interactive fusion on the corresponding deep semantic features and shallow texture features by adopting a deep and shallow multi-scale feature interaction mechanism; performing optimization by adopting cosine reconstruction loss, filtering consistency, entropy regularization loss and distribution alignment loss, and adaptively learning normal distribution characteristics according to an optimization process to obtain a model weight; and reasoning the input ultrasonic C-scan image by using the trained network weight to realize anomaly detection and positioning, and outputting an interpretable anomaly thermodynamic diagram.

The invention provides an automatic driving takeover quality evaluation method based on fusion of driving behaviors and physiological and psychological characteristics, and the method comprises the steps: carrying out the multi-source data collection and time synchronization of the behavior response and physiological change of a driver in the automatic driving takeover process, and obtaining multi-modal data; performing signal cleaning and adaptive segmented filtering on the multi-modal data to obtain filtering data; differential dynamic feature extraction and time window construction are performed on the filtering data, and multi-modal feature fusion is performed after key features are obtained, so that comprehensive features can be obtained; performing unsupervised clustering on the comprehensive features to obtain different takeover modes; and constructing an intermediary effect model based on the structural equation model and a Bootstrap test method, and performing quality evaluation on the takeover mode to obtain an evaluation result. According to the method, an intelligent evaluation framework combining Gaussian mixture model clustering and intermediary effect analysis is introduced, so that multi-dimensional and interpretable modeling and quality judgment of driver takeover behaviors are realized.

The invention discloses a boiler operation state intelligent monitoring method and system, and the method comprises the steps: collecting multi-source heterogeneous data in a boiler operation process, carrying out the cleaning and normalization processing of the multi-source heterogeneous data, extracting key features through wavelet transform, and obtaining standard multi-source heterogeneous data; modeling time series data in the standard multi-source heterogeneous data on the basis of an LSTM (Long Short-Term Memory) network, capturing a dynamic change trend of boiler operation to obtain time series characteristics, analyzing a hearth flame image and an infrared thermal image by using a CNN (Convolutional Neural Network), and extracting combustion state characteristics and thermal distribution characteristics; and based on the comprehensive state vector, a confidence interval of each feature dimension is calculated through a GMM Gaussian mixture model, an operation state deviation degree is analyzed and evaluated in combination with an entropy value, and when the deviation degree exceeds a preset threshold value, graded early warning is triggered. And the accuracy of boiler operation state intelligent monitoring is improved.

The invention discloses a multi-dimensional distribution offset adaptive threshold detection method and system, and the method comprises the steps: obtaining real-time data, extracting a multi-dimensional statistical feature, and obtaining a feature vector; based on historical normal data, using an expectation maximization algorithm to train a Gaussian mixture model, and determining parameters to obtain a normal distribution model; inputting the feature vector into the model, and calculating a probability value of the feature vector belonging to normal distribution as a first offset judgment index; based on the real-time data distribution of a plurality of detection objects in the same group, the distribution difference of any two objects is calculated by using a Wasserstein distance, and the similarity between the objects is obtained; and constructing a similarity network and calculating connectivity as a second offset judgment index. Setting a fixed-length sliding window, dynamically updating two indexes in the window, and obtaining a first self-adaptive threshold value and a second self-adaptive threshold value; and when any index is lower than a corresponding threshold value, determining distribution offset and giving an alarm, and updating model parameters in real time by using an incremental expectation maximization algorithm. According to the invention, accurate detection and intelligent analysis of data distribution offset are realized.

The invention discloses a cerebral apoplexy focus detection method and system, and belongs to the technical field of medical image detection. Extracting a fusion feature map of the brain image; for each region type, obtaining a representative feature which has the highest similarity with the feature at each pixel point position in the fused feature map in the class prototype set and carries a corresponding region type label, and further determining the region type to which each pixel point position in the fused feature map belongs so as to obtain a corresponding pseudo-label map; the class prototype set comprises representative features of different region types and is obtained in the training process of the system, and feature distribution of different region types in the memory bank is calculated through a Gaussian mixture model; for each region type, sampling is carried out based on feature distribution of the region type, and a plurality of representative features are obtained; the prototype-like set in the cerebral apoplexy detection method has real global context perception ability, can clearly distinguish the focus and various complex background structures, and can accurately realize cerebral apoplexy detection.

The invention discloses a fault diagnosis method and system for a rotating machine bearing, and belongs to the technical field of fault diagnosis. The method comprises the following steps: acquiring a to-be-diagnosed vibration signal of the rotary mechanical bearing; inputting the to-be-diagnosed vibration signal into a pre-constructed bearing fault diagnosis model, and outputting a fault diagnosis result of the rotating machine bearing; the bearing fault diagnosis model comprises a feature extraction network, a GMM feature enhancement module, a projection distillation module and a dynamic extensible classifier which are connected in sequence. According to the method, catastrophic forgetting can be effectively relieved, continuous diagnosis of newly-occurring accidental faults can be achieved, in the initial task, the model learns and identifies different fault types through a fault bearing data set collected in advance, a basis is provided for the subsequent increment task, the increment task is composed of a plurality of stages, and in each stage, the fault bearing data set is subjected to fault detection. The model enhances playback of old knowledge through historical category pseudo features generated by a Gaussian mixture model, and aligns feature representations of new and old models by using a projection distillation module.

The embodiment of the invention discloses a reachable probability calculation model training method, a control method, equipment and a storage medium. The training method comprises the following steps that according to design parameters of the mechanical arm, to-be-screened tail end pose data are determined; selecting reachable end pose data from the end pose data to be screened, and generating a training set; a multivariate Gaussian mixture model used for fitting reachable pose data space distribution is obtained based on training of the training set, a reachable probability calculation model is generated based on the multivariate Gaussian mixture model, and the reachable probability calculation model is used for outputting the reachable probability corresponding to any tail end pose. Therefore, the calculation model can be constructed and trained by adopting the architecture of the multivariate Gaussian mixture model. When control is carried out, control is carried out through model calculation of the reachable probability, the failure caused by the fact that the mechanical arm can be grabbed but cannot be reachable is avoided, and the operation success rate of the mechanical arm in the complex environment is increased.

The invention discloses a B5G base station T / R assembly health measurement method and system based on sparse projection and a hidden Markov model, and the method comprises the steps: collecting the multi-source operation data of a T / R assembly, and constructing a multi-dimensional operation parameter time sequence; a sparse projection matrix is constructed through covariance analysis and eigenvalue decomposition, and high-dimensional operation parameters are mapped into low-dimensional sparse health eigenvectors; training a hidden Markov model based on the feature sequence in the normal state, fitting an observation probability by using a Gaussian mixture model, and establishing a normal state reference model; calculating a KL distance between the current feature sequence distribution and the reference distribution, and mapping the KL distance into a normalized health degree index; and adaptively determining a state division threshold value by using a K-means clustering algorithm to realize health state grading of the T / R assembly. According to the method, environmental noise is effectively stripped through sparse projection, the dynamic reference model is utilized to adapt to complex working conditions, and online sensing and accurate evaluation of early weak degradation of the B5G base station assembly are realized.

The invention belongs to the technical field of adversarial samples. The invention provides a robust generalization-oriented few-sample continuous confrontation defense method. According to the embodiment of the invention, edge distance loss is resisted, in the pre-training stage, by maximizing the distance between the clean sample and the model decision boundary, the discrimination capability of the model for easily confused samples near the boundary is explicitly improved, the robustness of the model for the subsequent few-sample adversarial adaptation stage is enhanced, and the generalization capability of the clean sample and the adversarial sample is improved. Gaussian mixture model prototype playback is also provided, modeling is carried out on historical adversarial domain feature distribution by using the Gaussian mixture model, pseudo features are generated for knowledge playback, original adversarial samples do not need to be stored, and the robustness of the model in new and old adversarial domains is improved. Besides, by designing multi-domain balance loss, in multi-domain continuous adversarial training, updating is facilitated for most historical domains by constraining a model updating direction, inter-domain loss variance is reduced, and cross-domain balance optimization is realized.

The invention discloses a hyperspectral image classification method based on frequency domain denoising and element gradient correction, and the method comprises the following steps: carrying out the preprocessing of all hyperspectral image data, and dividing an overall training sample set formed by the processed hyperspectral images into a training set and a verification set; constructing a sample weighting model based on frequency domain denoising and element gradient correction; in the training process, a parameterization frequency spectrum gating sensing transformation module is utilized to map features to a frequency domain through discrete Fourier transform, a learnable frequency spectrum response function is utilized to adaptively suppress spectrum jitter noise, and finally pure features are reconstructed. And automatically constructing a high-confidence pseudo-clean verification set based on a Gaussian mixture model and time domain consistency. According to the method, a time domain momentum updating mechanism is introduced, the variance of statistical estimation is effectively smoothed, random interference caused by training fluctuation is resisted, and the accuracy of pseudo clean set construction and the convergence stability of overall model training are further improved.

The invention provides a big data advertisement putting method and system based on a cloud platform. The method comprises the following steps of performing multi-source data collection and preprocessing; dynamically constructing a user portrait; performing multi-dimensional analysis on the behavior pattern; target audience subdivision is based on a Gaussian mixture model; dynamically generating personalized advertisement creativity; a step of optimizing a multi-target delivery strategy; a step of real-time bidding and bidding optimization; a step of putting execution and real-time monitoring; a step of effect evaluation and multi-dimensional feedback; and carrying out adaptive system adjustment and iterative optimization. The big data advertisement putting method based on the cloud platform has the advantages that putting precision and efficiency are improved, audiences are subdivided through complex statistics, dynamic creativity is generated, advertisement correlation is enhanced, the predicted click rate is increased by 15-20%, and budget waste is reduced. And the cloud platform supports stream processing and automatic adjustment, so that the system can quickly respond to market changes, and the delay is reduced to a millisecond level.

The invention discloses an atmospheric greenhouse gas delta13C isotope observation construction method, which belongs to the technical field of atmospheric environment monitoring and is realized through three steps of monitoring data acquisition, traceability data acquisition and data processing standardization. According to the invention, through a double-site cooperative monitoring network and in combination with data acquisition of background sites and urban sites, the atmospheric CH4 concentration can be monitored in real time, and the delta 13C isotope value of CH4 can be accurately analyzed. And the source of CH4 can be qualitatively identified by utilizing a Kearing plot model, and high precision and high credibility of source analysis are ensured. According to the method, the Bayesian mixture model and the machine learning technology are adopted, the CH4 emission contributions of different sources are quantitatively estimated, and the emission sources and driving factors of CH4 are comprehensively analyzed in combination with multi-source tracer data and social economic data.

The invention belongs to the field of pollution traceability, and particularly relates to a karst landform soil and groundwater collaborative heavy metal pollution traceability method, which comprises the steps of end member site selection, groundwater monitoring point arrangement, collection and measurement of feature identification of each end member and multi-end member hybrid modeling. According to the scheme, the number of karst funnels is counted through neural network segmentation, high, medium and low density areas are divided, intelligent partition extraction of end members is achieved through a random forest model, monitoring points are arranged in the flow direction in a layered mode in combination with funnel distribution and an aquifer structure, and the problems that the pollution diffusion rule is difficult to capture and pollution source positioning is fuzzy under heterogeneity are solved; a three-dimensional monitoring network of soil, water quality and geology is constructed, a Bayesian mixture model is combined with an isotope fractionation effect to quantify an end member contribution proportion, the defect that traditional monitoring and modeling are not adaptive to karst heterogeneity is compensated, and pollution spatial and temporal distribution and migration paths are captured; unification of accurate simulation of the pollution diffusion rule and quantification of the end member contribution proportion is realized.

The invention relates to the technical field of multi-view clustering analysis, and discloses an incomplete multi-view clustering method and system, a storage medium and equipment. In order to solve the problems that a small amount of easy-to-obtain supervision information is not utilized in an existing method, and the clustering performance is deteriorated due to low view interpolation quality under high missing degree, the invention provides a technical scheme of combining paired constraint weighted interpolation and double-level feature fusion. The method comprises the following steps: firstly, screening similar samples by using pairwise constraint information, and reconstructing a missing view through similarity weighting; then extracting feature mean values and standard deviations of all views based on an encoder, aligning features through view hierarchy self-adaptive comparison learning and reserving private information, and optimizing feature distribution in combination with sample hierarchy semi-supervised loss; and finally, completing clustering through a Gaussian mixture model. According to the method, the pairwise constraint information is effectively utilized, the view recovery quality and the feature complementarity under the high-missing scene are improved, the clustering performance is remarkably improved, and the method is suitable for scenes such as data analysis of multi-view data missing and the like.

The invention belongs to the technical field of distributed multi-sensor passive detection multi-target tracking. The invention provides a factor graph optimization passive detection multi-target tracking method based on a Gaussian mixture model. According to the embodiment of the invention, the multi-target batch number is distributed by constructing the distributed passive sensor cooperative coordinate system and combining the multi-target identity judgment result between the two sensors; calculating direction finding lines based on two-dimensional observation of a sensor, combining the direction finding lines of the same batch number, obtaining a multi-target position estimation point set through a least square method, and obtaining a multi-target coarse positioning point through weighted fusion; modeling by adopting a Gaussian mixture model, fusing measurement distribution characteristics, solving parameters through an expectation maximization algorithm, and completing solvable conversion of an optimization problem; a factor graph optimization model containing multiple factors is constructed, state estimation is achieved through sliding window optimization, and track association and state updating are completed in combination with the Mahalanobis distance and the Hungary algorithm; and the passive detection multi-target tracking performance of the distributed sensor is effectively improved.

The invention relates to a power grid side energy storage two-stage locating and sizing method in a sea wind access scene considering the influence of temperature on a tide safety domain. The method comprises the following steps: adopting a temperature trend prediction model combining LSTM and TCN; a Gaussian mixture model is adopted to calculate a temperature variable capacitance expectation coefficient, and a power flow safety domain is reconstructed; establishing a random optimization scene set of source-load space-time coupling by adopting a typical source-load daily trajectory clustering method; establishing a power distribution network energy storage locating and sizing planning layer model taking the maximum comprehensive utility as a target; a power distribution network energy storage locating and sizing operation layer model with the minimum voltage fluctuation quadratic sum as the target is established; constructing a DistFlow multi-period optimal power flow model under the sea wind access scene; and solving the planning layer by adopting second-order cone relaxation, and carrying out secondary optimization on a charge-discharge and capacity configuration strategy on the operation layer by adopting a target cascade algorithm on the basis of anchoring a locating and sizing decision variable of the planning layer. According to the method, planning and operation targets can be more effectively coordinated, and comprehensive optimal configuration of power grid side energy storage is realized.

The invention discloses a crowd density detection method fusing optical flow and texture features, and relates to the technical field of computer vision, and the method comprises the following steps: collecting a real-time video stream of a camera, and carrying out graying, Gaussian filtering and perspective correction preprocessing; performing motion compensation by using image registration and offset transformation; modeling based on a Gaussian mixture model and extracting a foreground to generate a binary mask; analyzing a foreground coverage rate, an optical flow and texture features; pre-defining a multi-ROI and a density early warning standard; inputting the fusion features into a regression model and outputting initial density; dynamically calibrating and correcting the deviation; and generating a thermodynamic diagram superposition video to realize visualization. According to the invention, the optical flow and texture features are fused to improve density estimation precision, illumination resistance and dynamic background interference resistance; the dynamic calibration maintains long-term accuracy, and the dynamic ROI adapts to scene change; the thermodynamic diagram can quickly identify risks, is adaptive to multiple scenes, and meets real-time monitoring requirements.

The invention discloses a frequency domain adaptive clutter suppression method and system for a dual-polarization weather radar, and relates to the technical field of weather radar processing. According to the invention, pulse-level quality control and ground feature identification marking are carried out on radar original echoes; calculating power spectrums of the marking units, performing dynamic noise measurement and data quality grading, and screening out effective processing units; constructing a Gaussian mixture model of a power spectrum of an effective unit, and iteratively estimating model parameters through an expectation maximization algorithm to realize complete separation of a ground feature spectrum and a meteorological spectrum in a frequency domain; inverting single-channel parameters based on the separated meteorological spectrum; a cooperative spectrum separation strategy is adopted for horizontal and vertical polarization channels, and dual-polarization parameters such as differential reflectivity, correlation coefficients and differential phases are inverted based on the filtered dual-channel signals; according to the invention, high-precision separation of clutters and meteorological echoes is realized, the loss of meteorological signals is significantly reduced while ground features are effectively suppressed, and the accuracy and reliability of dual-polarization parameter inversion are improved.

The invention belongs to the technical field of data mining and artificial intelligence, and discloses a high-dimensional data clustering and feature structure analysis method, which comprises the steps of 1, preprocessing high-dimensional data and outputting standardized data to obtain a preprocessed standardized data set, 2, constructing a Gaussian graph mixture model to perform unsupervised clustering, and 3, obtaining a feature structure of the Gaussian graph mixture model; the method comprises the following steps: step 1, carrying out classification on the data sub-groups, and outputting data sub-groups containing class labels, step 3, independently constructing a feature association network for each class of data sub-groups, and outputting a topological graph of the feature association network of each sub-group, and step 4, carrying out multi-dimensional graph theory index calculation and statistical test on the feature association network of each sub-group, and outputting a final analysis result. According to the method, the problem that high-dimensional complex manifold distribution data is difficult to process is solved, core features behind different categories and a dynamic interaction mechanism thereof can be visually displayed, spanning from sample division to mechanism revealing is realized, and the method can be widely applied to the fields of medical subtype discovery, financial risk conduction analysis, industrial fault diagnosis and the like.

The invention relates to the technical field of image recognition, in particular to an abnormal security scene monitoring method and system for infrared thermal imaging, and the method comprises the steps: collecting a picture of a monitoring region through an infrared thermal imager, building a video frame sequence, and constructing a Gaussian mixture model for each pixel point; performing spatial analysis on the video frame sequence, establishing a foreground patch, and determining artificial environment spatial interference of each pixel point in the current frame through the foreground patch; performing time analysis on the video frame sequence in combination with the artificial environment space-time interference to obtain the artificial environment space-time interference of each pixel point in the current frame; evaluating space-time interference of an artificial environment by using a Gaussian mixture model, and judging and marking whether each pixel point in the current frame is in a candidate state or not; and presetting a decision threshold, counting the number of continuous frames with candidate state marks, comparing the number of continuous frames with the decision threshold, and distinguishing normal and abnormal environmental changes. Benign environment changes and real threats are effectively distinguished, and a large number of invalid alarms are prevented from being generated.

The invention discloses a traffic flow state data time gathering method based on Gaussian mixture probability, and belongs to the technical field of calculation, reckoning or counting. The method comprises the following steps: acquiring traffic flow state data and vehicle trajectory data of each lane in a target road section; carrying out probability diagnosis on the traffic flow data by adopting a Gaussian mixture model, and identifying and deleting abnormal points; introducing a model based on Transform and graph neural network fusion, and realizing space-time repair of abnormal data through joint modeling of an attention mechanism and a space-time topology dependency relationship; and carrying out time window collection on the repaired data based on wave walking and stopping prediction, and calculating a traffic flow rate, an average speed and an occupancy rate. According to the method, the joint characteristics of the traffic flow state data and the trajectory data are fully utilized, meanwhile, the abnormal data repairing precision and the space-time consistency are improved, adaptive collection of the traffic operation dynamic characteristics is achieved by introducing walking and stopping wave prediction, and the robustness and reliability of traffic flow data processing are remarkably improved.

The invention relates to the technical field of cross-domain facial expression recognition, in particular to a cross-domain facial expression recognition method and system based on intelligent learning. A shared backbone network is adopted to extract global depth features of the face image, and a coarse-fine branch network is constructed to realize collaborative learning of attitude analysis and emotional interpretation; a coarse and fine feature interaction mechanism is introduced in the training process to relieve subdivision branch too fast convergence, and uncertainty distribution of attitude analysis and emotional interpretation is obtained through similarity measurement and a Gaussian mixture model; modeling a potential feature linear dependency relationship, performing multi-source domain feature alignment, inhibiting tag noise and extracting robust semantic features; and performing cross-domain coupling calculation on the label uncertainty distribution and the semantic features to generate a final recognition result of the cross-domain facial expression. According to the method, the influence of cross-domain difference and label noise of expression recognition from coarse attitude analysis to fine emotion interpretation can be effectively relieved, and the generalization ability of cross-domain facial expression recognition and emotion interpretation precision are improved.

The invention relates to a structure vibration grading early warning method and device and a storage medium, and the method comprises the following steps: obtaining and preprocessing a structure vibration time sequence signal, firstly preprocessing the structure vibration time sequence signal, and dividing a data set; constructing a CNN-LSTM time sequence prediction model to carry out training optimization and short-term prediction; through historical data feature extraction and manifold dimensionality reduction, a vibration state mapping area is determined based on Gaussian mixture model clustering, and a posterior probability is calculated; a classification boundary is determined through maximum likelihood function iteration, and a multi-level early warning threshold value is established; and finally, three-level intelligent early warning of safety, warning and danger is realized by combining the dimension reduction point probability and the threshold value of the prediction data. Compared with the prior art, the intelligent multi-level early warning system is used for realizing high-dimensional time sequence prediction and dimension-reduction multi-level classification early warning in a high-efficiency and high-precision manner, and aims to establish the intelligent multi-level early warning system which can perform structure safety assessment more accurately and more timely.

The invention discloses an industrial network traffic anomaly detection and source tracing method, device and equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: converting a traffic data package of an industrial network into a process characteristic analysis software package, extracting a plurality of dimension statistical features from the process characteristic analysis software package, and storing the extracted statistical features in a database; converting the statistical characteristics of each dimension into table data; performing classification coding processing on connection behavior features in the table data to obtain a target coding result, normalizing statistical features to obtain a normalization result, converting time features to obtain a conversion result, obtaining a mixed feature vector, constructing a DAGMM model comprising a self-codec and a Gaussian mixture model by using an unsupervised mode, and obtaining a mixed feature vector; performing anomaly detection on the mixed feature vector to obtain an anomaly detection result; and if the traffic is abnormal, performing reverse decoding on the mixed feature vector by using a tracing algorithm to obtain a tracing result, thereby improving the efficiency of performing anomaly detection and tracing on the industrial network traffic.

The invention relates to a distribution network line loss probability distribution calculation method based on an improved Gaussian mixture model and linear power flow, and belongs to the field of distribution network line loss probability distribution calculation. According to the method, the optimized Gaussian mixture model is utilized to accurately describe the non-Gaussian characteristic and the multi-peak characteristic of photovoltaic output, and the defect that a traditional parameter estimation method is prone to falling into local optimum is overcome; the analysis mapping relation between the photovoltaic output and the system line loss is established through the linearization power flow model, a large amount of repeated calculation of the Monte Carlo method is avoided, the calculation efficiency is remarkably improved while the calculation precision is guaranteed, and the real-time requirement of engineering practice for rapid line loss analysis can be met.