911 results about "Single model" patented technology

The invention relates to the field of mechanical equipment health management, in particular to a large sliding bearing fault detection and evaluation method, device and system. Comprising the following steps: collecting multi-source sensing data, and constructing a comprehensive data set; constructing a state space model based on a sliding bearing physical mechanism; the multi-source sensing data and the state space model are fused through Bayesian filtering, and hidden state parameter posterior distribution is dynamically estimated; generating a virtual fault sample by using a generative adversarial network in combination with a physical rule base; designing a Bayesian space-time sequence diagnosis model based on an attention mechanism, and generating fusion health state features; processing and fusing the health state features by using a degradation process model, and predicting the remaining service life of the bearing; and based on the health state, the fault probability and the remaining service life, setting multi-stage early warning threshold values, and triggering intelligent early warning. According to the method, the defect that a single model is insufficient in adaptability and generalization ability under complex working conditions is overcome, and the accuracy and reliability of fault detection are remarkably improved.

The invention belongs to the technical field of image recognition, and provides a target tracking method and system based on AI vision, and the method comprises the following steps: collecting original video frames; environment adaptive image enhancement; performing multi-target detection and multi-modal feature extraction; estimating local optical flow motion; performing multi-target trajectory association; carrying out shielding processing and re-identification; outputting a track and analyzing a result; according to the method, a physical-deep learning cascade defogging model is set, light / dense fog processing paths are dynamically switched through a dark channel mean value, atmospheric scattering physical prior and U-Net residual error correction are fused, an environment self-adaptive sensing architecture is provided, the failure bottleneck of a traditional single model under sudden change fog concentration is broken through, and the real-time performance of the system is improved. According to the method, an apparent-motion-geometry ternary coupling trajectory cognition system is constructed, a dynamic cost matrix and a feature cache pool are designed, the ID switching problem caused by similar target aggregation and long-time shielding is solved, and the accuracy of target tracking in the shielding environment is improved.

The invention provides a wind power prediction method and system based on time sequence decomposition and multi-model fusion, and the method comprises the steps: collecting the historical power generation power and meteorological data of a target wind power plant; decomposing the historical power generation power and the meteorological data to obtain a trend component, a seasonal component and a residual component; fusing with meteorological data to construct a trend feature matrix, a periodic feature matrix and a residual feature matrix; different modeling schemes are adopted to construct corresponding single models; dividing into a training set, a verification set and a test set according to a time sequence; performing training optimization on the single model by using the training set, the verification set and the test set, and constructing a wind power short-term power prediction model; and inputting the real-time meteorological data and the generated power to the wind power short-term power prediction model, and outputting the generated power prediction value of the target wind power plant, thereby effectively improving the comprehensiveness, accuracy and stability of model prediction.

The invention discloses a storm surge water increase prediction method and device, electronic equipment and a storage medium, and relates to the technical field of seawater monitoring, and the method comprises the steps: obtaining multi-source observation data of a target sea area, the multi-source observation data comprising satellite remote sensing data, near-shore monitoring data, meteorological mode data and drainage basin data; performing space-time alignment and exception processing on the multi-source observation data to generate standardized space-time grid data; performing hybrid prediction modeling on the standardized space-time grid data through a physical data dual-drive modeling layer to obtain a target probability water increasing field; and dynamically correcting the target probability water increasing field by using real-time observation data to obtain a storm surge water increasing predicted value. A physical mechanism model and a probability generation model are combined, limitation of a single model is overcome, and the nonlinear evolution process of the storm surge is effectively captured; the time-space dual-drive architecture is adapted to complex coast terrains and changeable meteorological conditions, and the risk of missing report or false report is reduced.

The invention discloses a maritime accident prediction method and device based on interpretable integrated machine learning, and relates to the technical field of maritime affair safety risk analysis, and the method comprises the steps: obtaining accident investigation data, carrying out the preprocessing, balancing the data through a ten-fold layered oversampling method, and carrying out the cross verification training, and determining a performance optimal model by using the test set and carrying out interpretable analysis to explain the influence of the characteristics on the accident prediction result. By constructing a closed-loop'data processing-model optimization-explanation output 'process and adopting SMOTE oversampling and ten-fold layered cross validation training and a heterogeneous base model ensemble learning strategy, the processing capacity of the data imbalance problem of accident categories is improved, the data leakage problem of oversampling is avoided, and the possible bias of a single model is overcome. The interpretability analysis of the model prediction result can quantitatively display the contribution degree of each feature to prediction globally and locally, reveal the nonlinear relationship and interaction effect between the features, and provide transparent interpretation of model decision.

The invention discloses an agent reasoning method based on a knowledge graph, and belongs to the technical field of knowledge graphs, and the method comprises the steps: integrating user query, the knowledge graph and multi-modal data through an NLP model and an entity linking technology, and extracting a target entity, a relation constraint and a structured feature vector. The LLM can synthesize more information to generate a more comprehensive conclusion, the knowledge embedding model maps an entity relationship into a geometric relationship in a vector space, the LLM is assisted to verify reasonability of reasoning, the system can comprehensively generate confidence through the LLM output probability, knowledge embedding similarity and data quality, and the reliability of a result can be explained through confidence score. The LLM generation capability is combined with the vector reasoning capability of knowledge embedding, and the limitation of a single model is made up.

The invention provides a spatial-temporal feature fusion-based dynamic graph convolution electroencephalogram depression detection method, which comprises the following steps of: firstly, segmenting a sample into fragments with the length of 1 second, and calculating power spectral density (PSD) as an input feature by adopting a Welch method; the time sequence and spatial characteristics of the EEG signals are synchronously extracted through a double-branch architecture, wherein one branch captures the long-term time sequence dependence of the EEG signals by using a GRU; and the other branch adopts an improved TSCN (separable convolution is introduced), multi-scale spatial features from fine to rough are extracted through causal convolution and expansion convolution of residual layered stacking, after double-branch features are adaptively fused based on an attention mechanism, a dynamic graph structure is constructed, functional connection evolution of brain intervals is modeled by using a graph convolution network, and a dynamic graph structure is constructed. The topological structure of the network is optimized through a back propagation process, and finally depression identification is realized through a Softmax classifier. According to the method, the time sequence modeling capability of the GRU and the multi-scale spatial analysis capability of the TSCN are fused, the representation limitation of a single model is broken through, the dynamic change of a brain function network is adaptively captured through dynamic graph convolution, the physiological interpretability is enhanced, deep complementary fusion of EEG spatial and temporal characteristics is realized, the depression recognition accuracy is remarkably improved, and the method is suitable for popularization and application. And an efficient tool is provided for auxiliary diagnosis of mental diseases.

The invention discloses a system and method for estimating the charging capacity of an energy storage lithium battery, and particularly relates to the technical field of energy storage battery management, and the method comprises the following steps: collecting dynamic parameters in the charging process of the energy storage lithium battery, and forming a dynamic data set; calculating an environmental disturbance influence index and a charging response consistency factor based on time sequence analysis and feature decoupling; constructing a two-dimensional working condition mapping matrix, identifying a current working condition area and generating a scene label; determining parameter weights of the plurality of estimation models by adopting a probabilistic reasoning mode; according to the scene state, selecting a single model output result or fusing a plurality of model output results for estimation; according to the method, the environment disturbance influence index and the charging response consistency factor are constructed, so that the complex working condition is accurately identified; scene labels are automatically generated based on two-dimensional working condition mapping and a clustering algorithm, and a plurality of estimation models are dynamically selected or fused in combination with model confidence, so that the accuracy, robustness and intelligent level of estimation are improved.

The invention relates to the crossing field of hydraulic engineering hydrological monitoring technology and machine learning prediction technology, discloses a sediment concentration prediction method based on deep learning, and aims to solve the problems that in existing sediment concentration prediction, hyper-parameter manual tuning is low in efficiency, key feature attention is insufficient, local and time sequence information is difficult to consider by a single model and the like. Accurate prediction is realized through five core modules: a data preprocessing module performs missing value filling, abnormal value processing and derivative feature generation on hydrological data; the feature selection module screens key features based on mutual information; the time sequence construction module generates time sequence data through a sliding window; the hyper-parameter automatic optimization module adopts Bayesian optimization iteration to obtain an optimal hyper-parameter; the CNN-LSTM-attention prediction module fuses CNN local feature extraction, bidirectional LSTM time sequence dependence capture and multi-head self-attention mechanism key feature focusing capability, is suitable for scenes such as river channels and channels, and provides efficient decision support for hydrological regulation and control.

The invention discloses a deep learning and machine learning fused soil moisture inversion construction method, and relates to the technical field of measurement of physical properties of materials, and the method comprises the steps: capturing complementary information and spatial context of multi-source data through a multi-source heterogeneous data space-time adaptive fusion step by using a cross-modal attention mechanism and a graph neural network; through a deep learning and machine learning dual-path collaborative inversion step, advantage complementation is realized by combining data-driven nonlinear modeling and a physical constraint interpretable model; according to the method, the defects of single data source, insufficient model generalization ability and incomplete physical mechanism consideration in the prior art are overcome, the inversion precision is improved by 12%-18% under the complex earth surface condition, and the method has the advantages that the method is suitable for large-scale popularization and application. And a high-precision, strong-generalization and reliable technical means is provided for precise monitoring of soil moisture.

The invention belongs to the technical field of geometric pretreatment in computer aided engineering (CAE), and particularly relates to a dirty geometry restoration method for a complex three-dimensional geometric model, which comprises a dirty geometry detection and restoration module based on heuristic rules and geometric topology analysis. By means of a multi-model collaborative alignment and bonding mechanism, high-precision assembly between parts is achieved through geometric feature matching and adjacent domain self-adaptive adsorption; acquiring material attributes provided by the simulation platform, and performing accurate cutting and boundary reconstruction on an intersection domain based on an intersection domain intelligent segmentation algorithm of the material attributes to ensure correct mapping of boundary conditions of a physical field; a geometric Boolean processing engine iterative learning mechanism of a general merging algorithm is integrated, and a high-quality model for simulating a grid is generated through multi-round iterative optimization. According to the method, double challenges of internal dirty geometry of a single model and derivative dirty geometry of intersection / adjacent areas among multiple bodies are solved, and key physical attributes and topological continuity of electromagnetic signal links are ensured and are completely reserved during repair.

The invention relates to the technical field of geological disaster monitoring, and discloses a landslide dynamic simulation monitoring and early warning method based on multi-source sensing fusion, and the method comprises the steps: collecting multi-source data of a landslide body through a plurality of heterogeneous sensors, and enabling the multi-source data to be in space-time alignment after preprocessing; building a multi-parameter fusion model fusing a displacement field, a mechanical field and an environment field based on the preprocessed data, enabling the multi-parameter fusion model to output a deformation rate and a stability coefficient, and dynamically adjusting the weights of the displacement field, the mechanical field and the environment field according to a landslide evolution stage; predicting a future deformation trend of the landslide mass in combination with a geological structure and historical data, comparing a stability coefficient with a dynamic safety threshold to judge a risk level, and generating early warning information; and dynamically correcting a reference weight coefficient in the model based on the deviation between the monitoring data and the model output, so that the model is adaptively optimized. The problems of single monitoring dimension and static model solidification in the prior art are solved, and accurate and adaptive monitoring and early warning of the risk state of the landslide mass are realized.

The invention discloses a resume analysis method and system based on multiple large language models, and belongs to the technical field of large language models.The resume analysis method and system based on the multiple large language models.The resume analysis method and system based on the multiple large language models comprise the following specific steps that firstly, model parallel analysis is conducted, simultaneously inputting the data into at least two heterogeneous large language models through an application program interface; and each large language model independently performs information extraction and analysis according to the model structure and the training data of the large language model, and outputs a structured data result containing a plurality of preset fields. Through the multi-model parallel analysis and conflict re-judgment mechanism, the misjudgment risk of a single model is effectively reduced, the robustness of the whole system is improved, the resume analysis accuracy is remarkably improved, the model pool is automatically optimized and updated through the dynamic scoring mechanism, and the problem that the model is difficult to select and update is solved.

The embodiment of the invention discloses a collaborative reasoning method and device based on a multi-expert model and a medium, and relates to the technical field of collaborative reasoning, and the method comprises the steps: receiving a plurality of reasoning task requests, so as to obtain the task feature information of each reasoning task; based on the task feature information of each reasoning task and pre-acquired model registration information, performing task allocation to determine at least one allocation expert model corresponding to each reasoning task; predicting the model load of each distributed expert model, adjusting at least one distributed expert model corresponding to each reasoning task according to the predicted load rate, and determining at least one corresponding target expert model; and issuing each reasoning task to the corresponding target expert model, collecting single model reasoning data of each target expert model through an asynchronous message queue, and determining a collaborative reasoning result corresponding to each reasoning task through the single model reasoning data of each target expert model.

The invention relates to a reservoir porosity and permeability prediction method based on a dynamic committee integration model, and the method comprises the following steps: obtaining an original data set which comprises shale content, porosity, permeability, GR, AC, CNL, DEN, RT, RXO, SP, CALC, CALI, depth and lithologic labels; performing data enhancement on the original data set; determining main control factors influencing the porosity and the permeability; constructing a dynamic committee integration model; inputting the main control factors into a dynamic committee integration model; and utilizing the trained dynamic committee integration model to respectively predict the porosity and the permeability. According to the method, the logging data is processed by adopting a machine learning method, reservoir parameters can be efficiently and accurately predicted, and favorable support can be provided for oil-gas exploration and development; the dynamic committee integrated model constructed by the invention can dynamically adjust the weight of each model according to different geological conditions and data features, can more flexibly adapt to the geological condition of a research area compared with a single model, and improves the prediction precision and generalization of the model.

The invention discloses a knowledge graph completion method based on combined prediction of large and small models, which comprises the following steps: 1, constructing and preprocessing a knowledge graph completion reference data set, and training by adopting a RotatE model to obtain candidate entities generated by the model and confidence scores; 2, constructing a related triad, an adjacent triad and entity long text description based on the query to form a context prompt, inputting the context prompt and the query into a large language model, and performing semantic reordering and scoring by the large language model; and 3, constructing a fine tuning data set, and performing fine tuning on the large language model to obtain the KGC task optimization-oriented large language model. And based on the KGC score, the LLM score and the dynamic weight, outputting a complementation result through joint prediction of a fusion result. According to the method, the structured reasoning ability of the small model and the deep semantic understanding of the large language are effectively combined, the prediction accuracy, robustness and specialty are remarkably improved, and the defects that a single model is weak in generalization ability and insufficient in semantic utilization are overcome.

The invention belongs to the technical field of gas emission concentration monitoring, and particularly discloses an Informer-based method for monitoring the emission concentration of harmful gas in a large-scale farm in real time. According to the method, aiming at the characteristics of high nonlinearity, variable redundancy, complex coupling and the like of the emission data of the farm, a fusion model based on an Informer model and an RLS model is constructed, and the perception and response capability to the harmful gas emission concentration change is improved. According to the fusion model, global trend prediction is achieved through an Informer model, then dynamic error correction is conducted through an RLS model, the real-time performance, self-adaptability and robustness of the fusion model are effectively enhanced in combination with a processing mechanism of prediction first and parameter lagging update, and the fusion model is obviously superior to a traditional static or single model prediction mode. The method provided by the invention effectively solves the problems of response lag and inaccurate prediction in the current harmful gas emission concentration detection technology.

The invention relates to the technical field of disaster prevention and reduction of a power system, and discloses an icing risk early warning method based on multi-model fusion and residual time sequence characteristic analysis, which comprises the following steps: collecting meteorological data of a line area in real time, removing abnormal values through secondary judgment of a Pauta criterion and a trend, and standardizing; adopting a TEROL algorithm to screen high-weight key features; running SWD-BP, MUL-GRNN and ELM models in parallel, constructing a dynamic weight by combining DSI, an independence weight method and an entropy weight method, and calculating a final meteorological predicted value; generating a prediction residual signal, extracting time domain features such as a mean value and a peak value, and constructing a residual feature matrix through a sliding window; and inputting an LSTM model to process a time sequence dependency relationship, and judging an icing risk level. According to the method, meteorological prediction is optimized through multi-model dynamic fusion, and deviation is analyzed and corrected in combination with residual time sequence characteristics, so that the problem of weak generalization ability of a single model is effectively solved, and the accuracy of icing risk early warning is obviously improved.

The invention relates to the technical field of traffic transportation environment engineering and carbon emission evaluation, in particular to a construction method and equipment of a plateau gradient zone vehicle CO2 emission factor calculation model and a medium. Through the method, a route-level CO2 emission factor distribution map is obtained, and different quantitative calculation models are formed for different altitude intervals. And a specific and quantitative design basis is provided for low-carbon highway design. It is revealed for the first time that the CO2 emission rate of the plateau gradient zone has a remarkable altitude segmentation effect through empirical data, and a method for constructing a calculation model in a partition mode is innovatively provided. According to the method, the defects that a traditional single model is poor in adaptability and misaccurate in prediction in a complex plateau environment are overcome, and differentiated quantitative models (such as a low-altitude quadratic function and a high-altitude S-type function) are established for different altitude intervals; the finally formed partition model provides a core theoretical tool and decision basis for plateau highway carbon emission accurate accounting, low-carbon route optimization and green traffic construction.

The invention relates to the technical field of genome data analysis, and discloses an intelligent system, method and equipment for assisting multi-step genome data analysis, and the system comprises a dialogue agent which is used for generating a corresponding answer according to a question of a user, or reading an analysis plan file generated by a workflow agent, generating an analysis interpretation text for the analysis plan file; the workflow agent is used for generating a structured task execution plan according to the to-be-executed analysis task and executing the to-be-executed analysis task; and the modeling analysis agent is used for generating a configuration file and a script based on the user request, constructing a model and generating an analysis result corresponding to the user request in combination with the workflow agent. Through multi-agent cooperation, task division and cooperative scheduling are realized, each agent independently completes task planning, execution control, model analysis and other functions, the bottleneck problem of processing of a traditional single model in a complex process is avoided, error accumulation is reduced, and the execution efficiency and stability of the whole process are improved.

The invention provides a lithium ion battery health state estimation method and system based on a dual-drive interpretable integrated model, and belongs to the field of lithium ion battery health state estimation. The problems that an existing lithium battery health state monitoring method is single in feature source, insufficient in model generalization ability and poor in interpretability are solved. According to the method, based on an incremental capacity curve and a first-order RC equivalent circuit model, IC peak value features and ohmic internal resistance features are extracted, and a multi-source health feature space is constructed in combination with voltage statistical features; an integrated learning model based on Stacking is established, three basic models of a random forest, kernel ridge regression and an interpretable enhancement machine are integrated, and collaborative optimization of model hyper-parameters is realized by adopting a tree structure-based Bayesian optimization algorithm.

The invention discloses a multi-degraded image recovery method based on frequency domain decomposition, and aims to solve the problems that a single model is difficult to deal with various image degradation and recovery processes of different frequency domains are mutually coupled in the prior art. According to the method, a degraded image is decomposed into a high-frequency space and a low-frequency space through fast Fourier transform, and a double-branch network architecture is adopted for targeted processing: for the high-frequency part, a high-frequency feature adaptive processing module HFPM is designed, and detail texture features are effectively extracted and interference is suppressed through feature enhancement and cross-layer fusion technologies; and for the low-frequency part, constructing a low-frequency feature conversion enhancement module LTEM, and capturing global context information by using cyclic convolution to improve the integrity of the structure contour. According to the method, decoupling processing of frequency domain features is realized, and the image restoration performance of the model in various degradation scenes such as rain removal, noise removal and defogging is remarkably improved through the synergistic effect of high-frequency detail enhancement and low-frequency structure optimization. Experimental results show that the method has excellent recovery effect and robustness when a plurality of image degradation tasks are processed at the same time, and can be effectively applied to visual tasks such as traffic accidents with high image quality requirements.

The invention relates to the technical field of automatic driving and artificial intelligence, in particular to a driving track planning method based on a visual language model and an intelligent driving system. The trajectory planning method comprises the steps of multi-modal data acquisition and preprocessing, visual language model reasoning, execution control and the like. The intelligent driving system comprises a multi-modal data acquisition and preprocessing module, a visual language model reasoning module and an execution control module. The visual language model reasoning module is integrated with a Patch selection module, a token compression module and a speculation decoding module; according to the invention, an end-to-end driving planning reasoning framework is constructed based on the visual language model, the vehicle trajectory planning is directly generated from the original multi-view perception data and the high-level driving intention by a single model, the architecture of an automatic driving planning system is simplified, and the response speed and robustness are improved. The method can effectively meet the requirement of the automatic driving vehicle for generating the safe path in real time in a vehicle-mounted embedded environment, and has important practical value.

The invention discloses a construction method of an intelligent question answering system based on a lightweight large model, and relates to the technical field of natural language processing, and the method comprises the steps: receiving a natural language question of a user, carrying out vector coding through a lightweight BERT model, calculating the cosine similarity of the natural language question and a business index vector, and generating a structured semantic map; converting the structured semantic map into a scene feature vector, and injecting the scene feature vector into an adapter parameter block to construct a lightweight scene adaptation model; the lightweight scene adaptation model is combined with a real-time data interface to obtain a structured multi-modal response and construct an index tracking tree; and based on the unexpanded nodes of the index tracking tree, actively initiating scene migration type questions through a questioning strategy engine, obtaining target scene feature vectors, performing deep analysis, and generating a deep question and answer analysis report. According to the method, the lightweight scene adaptation model is constructed, calculation logic is flexibly adjusted according to different service scene features, and the self-adaptive processing capability of a single model to multiple service scenes is achieved.

The invention belongs to the technical field of sound source processing, and provides a sound source positioning and detecting method and device. The method comprises the following steps: acquiring delay estimation of microphones I and II and delay estimation of microphones I and III based on a three-path linear uniform microphone array; based on the delay estimation sum, respectively carrying out positioning calculation under near-field and far-field conditions; according to the sound source distance under the near-field condition, comparing the sound source distance with a distance judgment threshold value, and determining a far-field / near-field output sound source position; and carrying out feature extraction on the signals of any microphone array, and carrying out event classification based on a pre-trained convolutional neural network. According to the method, far / near field model selection is carried out according to the distance judgment threshold, large deviation generated by a single model in a critical region is avoided, continuous and stable positioning from short distance to long distance is achieved, time classification can be achieved while position calculation is carried out, and integrated output is achieved.

The invention provides a civil aviation pilot on-duty physiological early warning method based on LSTM and GBDT. The civil aviation pilot on-duty physiological early warning method comprises the following steps that S1, multi-source physiological data and associated information are collected; s2, data preprocessing and quality verification; s3, carrying out engineering construction on the physiological features and the associated features; s4, training and optimizing an anomaly recognition model; s5, carrying out real-time physiological anomaly recognition before attendance; s6, multi-level early warning triggering and information pushing are carried out; s7, performing early warning feedback and model iterative optimization; and S8, performing abnormal event tracing and data closed-loop updating. The limitation of a single model is overcome through multi-algorithm fusion, and the recognition accuracy in a complex scene is improved; a hierarchical early warning and closed loop optimization mechanism is established, safety guarantee and operation efficiency are considered, a standardized and landing technical scheme can be provided for civil aviation pilot health management system construction, and reduction of flight safety risks caused by physiological abnormalities is assisted.

The invention discloses a total primary productivity estimation method and system based on multi-model coupling deep learning, and the method comprises the steps: obtaining the multi-source data of meteorological data, remote sensing images, latent heat flux, sensible heat flux and solar radiation, carrying out the quality control, missing value processing and nearest neighbor interpolation of different data sources, and carrying out the prediction of the total primary productivity. Unifying to a target spatial resolution and a time resolution; in a light energy utilization rate (LUE) model family, a solar radiation phase factor is introduced into a photosynthetically active radiation absorption ratio (FPAR) to obtain a phase modulation type FPAR, drought duration is introduced into a water stress function f (W) to obtain an exponential decay type f (W), and a GPP time sequence of a plurality of improved mechanism models is calculated according to the exponential decay type f (W); extracting spatial texture features from a remote sensing image stack by using a convolutional neural network (CNN), and performing cross-modal fusion on the spatial features and the GPP estimated by the plurality of improved mechanism models in a gating mode to form fusion representation; carrying out learning and collaborative optimization on the fusion representation of the GPP and CNN spatial features estimated by the plurality of improved mechanism models by adopting a gradient lifting tree model; and high-precision estimation of the GPP is realized through multi-model collaborative optimization. The method aims at solving the problem that a traditional light energy utilization rate model is insufficient in response under the extreme environment conditions of drought and intense radiation, the adaptability limitation of a traditional single model under the complex environment is broken through, and therefore high-precision GPP estimation under the complex environment is achieved.

The invention discloses an interpretable false news detection method and system based on retrieval evidence enhancement, the interpretable false news detection method combines a big language model with an external knowledge base retrieval mechanism, that is, factual supplementary evidence is provided for the big language model to generate final reasoning through a retrieval module. Screening candidate text evidences by utilizing a clustering algorithm, and performing credibility scoring and filtering on the candidate evidences in combination with a large language model; a unified reasoning Prompt is constructed to input a plurality of large language models, true and false judgment is output respectively, and detailed explanatory texts are generated with the assistance of natural language explanation, so that the false news detection method with factual support, language expression and user understandability is realized. And the judgment results of the models are fused through a majority voting mechanism, and explanation is generated from multiple perspectives, so that the risk caused by reasoning deviation of a single model is reduced, and the accuracy and stability of system output are effectively improved.

The invention provides a modal unified data processing method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence, in particular to the technical fields of deep learning, large models and the like. The specific implementation scheme is as follows: determining a plurality of first weight parameters corresponding to a plurality of attention experts in an attention layer by using a first gating network according to vector features of input lexical elements; based on the plurality of first weight parameters, processing the input lexical element by using at least one activated attention expert in the plurality of attention experts to obtain an attention layer output vector; using the first residual connection and the normalization layer to generate an intermediate lexical element according to the attention layer output vector and the input lexical element; and processing the intermediate lexical elements by using the feed-forward network layer to generate output lexical elements. According to the scheme of the embodiment of the invention, intermodal interference can be effectively relieved, and a single model can flexibly and efficiently process multi-modal and multi-task data.

The present invention belongs to the field of image classification and designs a long-tail image data classification method based on mixed samples. The present invention proposes a new solution to the long-tail training set problem encountered in image classification research. It aims to use three experts with specialized knowledge to jointly assist the algorithm in making the final decision, avoiding problems such as excessive deviation of the model classifier weight caused by a single model. The present invention is suitable for business scenarios of image classification with long-tail data distribution. By designing multiple experts with specific field knowledge, the classification performance of the model for all frequency distribution types is improved without losing the accuracy of the head class classification. It provides a solution for the actual engineering application of image classification when the data has a long-tail distribution, alleviates problems such as data collection difficulties, improves the overfitting of the algorithm model to the head class data, and improves the learning ability of the tail class data.