172 results about "Multiple classification" patented technology

The invention discloses a node classification method and system based on a hybrid expert model.The node classification method comprises the steps that node data are input into a pre-trained hybrid expert model to obtain a classification result of the node data, and the hybrid expert model comprises a node mode extractor, a gating model and a plurality of classification models; the node mode extractor is used for extracting node mode information from given node data, and the gating model is used for receiving the node mode information and generating a model combination weight corresponding to each node; and dynamically selecting and combining the classification results of the node data by the plurality of different types of classification models according to the model combination weights corresponding to the nodes so as to obtain the final classification result of the nodes. The invention aims to solve the problem of node classification under a complex graph structure, accurately define a node mode under the condition that the node mode is complex, and properly combine different node predictors for different nodes.

The system and method for processing a digital pathology image using a machine learning model that includes a self-supervised hierarchical Vision Transformer (ViT) configured to perform unsupervised clustering with multiple classification tokens. The method includes receiving a digital pathology image that depicts a tissue slice stained with histological dyes. The digital pathology image may be processed to generate a result comprising multiple predicted classifications of individual patches of the digital pathology image. The result is generated by a machine-learning model using a self-supervised hierarchical Vision Transformer (ViT) that may further comprise a multi-head self-attention module configured to predict a crosspatch relevance metric using an attention mechanism for each individual patch in the digital pathology image thereby assigning the individual patches to a cluster based on the crosspatch relevance metrics.

The invention discloses a text classification method and device, a nonvolatile storage medium and electronic equipment. The method comprises the following steps: determining a text feature vector of a to-be-processed text and task types of a plurality of classification tasks; a multi-task processing model is adopted to process the text feature vectors to obtain task feature vectors corresponding to the multiple classification tasks, the multi-task processing model comprises multiple feature extraction layers, and each feature extraction layer is provided with a first gating network used for controlling output feature weights of the shared expert network and a second gating network used for controlling output feature weights of the shared expert network; and a second gating network for controlling output feature weights of the task specific expert network; and processing the task feature vector corresponding to the classification task through a classifier corresponding to the classification task to obtain a classification result of the classification task. According to the method and the device, the technical problem that the final classification result is low in accuracy due to the fact that one gating network is adopted to control the output feature weights of the shared expert network and the specific expert network at the same time in the related technology is solved.

Methods and systems are provided for improving substrate defect classification in semiconductor manufacturing by using more than one machine learning model to classify substrate defect data. The method comprises a defect inspection module that captures substrate defect image data and a defect classification part that processes the data using more than one machine learning model. The output from each model is used to produce the final classified data. The defect score is calculated based on the classification results and this defect score is used to identify the misclassified substrate defect. The model can be updated after each inspection run cycle leading to increased accuracy and a lower escape rate.

The invention discloses a cancer classification comparison learning method based on a twinborn network, and the method comprises the steps: obtaining a plurality of sample pathological section images and test pathological section images through collection equipment, carrying out the cancer type judgment of each sample pathological section image, carrying out the marking according to a judgment result, obtaining corresponding first marking data, and carrying out the classification comparison of the test pathological section images; the method comprises the following steps: respectively carrying out similar and heterogeneous random pairing on different sample pathological section images according to cancer categories to obtain a plurality of training pairs, creating a twin network, sequentially inputting each training pair into the twin network, and respectively learning similarity measurement between the pathological section images in different training pairs to obtain a trained twin network, and acquiring and using feature vectors with representation capability generated during twin network training and corresponding second annotation data to train a cancer classification prediction model, inputting all the test pathological section images into the cancer classification prediction model to obtain a plurality of classification prediction results, and visualizing the classification prediction results.

The invention discloses an order dispatching method and device, a nonvolatile storage medium and electronic equipment. The method comprises the steps that a to-be-dispatched fault work order is acquired, and complaint information received in a preset time period is recorded in the to-be-dispatched fault work order; the work order classification model is called to perform content analysis processing on the to-be-dispatched fault work orders to obtain a classification result output by the work order classification model, the classification result is used for indicating a dispatch result corresponding to each to-be-dispatched fault work order, and the work order classification model is trained by adopting historical complaint information and historical dispatch results as training data; feedback results of the multiple classification results are obtained, and the feedback results comprise verification results and correction results of the classification results; and according to a feedback result, carrying out order sending operation. According to the invention, the technical problem of low order sending accuracy caused by manual order sending is solved.

The invention relates to a method for training machine learning models, having the steps of: detecting data in the form of time series data using one or more computers, said data being obtained by means of one or more measuring devices (60-62), in each case in the form of a sensor for measuring a physical variable; receiving multiple classification data units relating to the data using the one or more computers; receiving a selected part of the data using the one or more computers for each of the classification data units; and training multiple machine learning models using the one or more computers, in each case on the basis of at least one of the classification data units and the at least one corresponding selected part of the data, wherein the multiple machine learning models represent multiple instances of the same machine learning model.

The present invention discloses a method and device for optimizing drilling parameters, wherein the method comprises: constructing a training set and a validation set; training a machine learning model according to the training set to obtain a plurality of classification prediction models respectively established for each formation; using a plurality of classification prediction models to obtain the prediction accuracy of each classification prediction model; selecting a first classification prediction model from each classification prediction model according to the prediction accuracy of each classification prediction model and a preset performance evaluation index; obtaining a data group that can achieve drilling speed-up during drilling based on the first classification prediction model, and establishing a random forest regression model; constructing a dual-objective evaluation function using the importance of characteristic parameters of the input data of the random forest regression model as weights; and determining the value range of each characteristic parameter according to the dual-objective evaluation function. The present invention can optimize the value range of drilling parameters that can achieve drilling speed-up.

The embodiment of the invention provides a multi-party joint model training method and device under non-independent identical distribution. And a plurality of participants respectively execute training and updating of the local model, send updated parameters to the central server, and perform parameter aggregation through the central server. Any participant constructs a reference probability of multiple classifications of the sample based on the label classification of any sample and the prediction probability of the local model for the multiple classifications of the sample, so that the prediction probability of the label classification is equal to the reference probability; the reference probability of the non-tag classification is set to the same value related to the predicted probability of the tag classification. A first loss is then determined based on a difference between the predicted probability and the reference probability for the number of non-label classifications. Meanwhile, a first conditional probability of the non-label classification is determined based on a prediction probability of the non-label classification of the samples of which the label classification is the first classification, a second conditional probability of the non-label classification is determined based on a prediction probability of the non-label classification of the samples of which the label classification is the second classification, and a second loss is determined based on a difference between the two conditional probabilities. And updating the local model in the direction of reducing the two losses. In the process that multiple parties jointly carry out model training, each participant needs to carry out privacy protection on privacy data of the participant.

The invention discloses an infant cry detection method, device and system and a storage medium. The detection method comprises the following steps: acquiring an audio sampling signal; processing the audio sampled signal to generate a logarithmic Mel spectrum feature; the logarithmic Mel spectrum features are input into a trained neural network algorithm model to be detected, a cry detection result is obtained, the detection model comprises a convolutional neural network algorithm, a prediction function and a judgment threshold value, the judgment threshold value can be adjusted, the convolutional neural network algorithm generates a multi-classification task result according to the logarithmic Mel spectrum features, and the multi-classification task result is subjected to multi-classification. The prediction function determines a classification result according to a multi-classification task result, and the judgment threshold is used for comparing with the classification result to obtain a cry detection result. According to the detection method, the input logarithmic Mel spectrum features are detected through the detection model to generate the multi-classification task result, and the cry detection result is obtained by comparing the multi-classification task result with the adjustable threshold value, so that the problem of how to balance false alarms and missing alarms is solved.

The invention discloses a self-adaptive dynamic data cleaning method for real-time streaming data, which relates to the field of data cleaning and comprises the following steps of: acquiring corresponding service demand data and real-time streaming data; classifying the obtained real-time streaming data for multiple times according to the service demand data, and inputting the real-time streaming data into corresponding data link channels; performing data anomaly evaluation on the corresponding real-time streaming data according to the classification processing result corresponding to the corresponding cleaning node in the data link channel to obtain corresponding node abnormal data, and setting a cleaning path of the corresponding real-time streaming data according to the node abnormal data; performing data cleaning on the real-time streaming data according to the corresponding cleaning path, performing real-time feedback on the cleaning process, and generating feedback sampling adjustment information; dynamically adjusting the acquisition process of the corresponding real-time streaming data according to the feedback sampling adjustment information, and outputting the real-time streaming data to complete the data cleaning process; the efficiency in the data cleaning process is improved to a certain extent.

Provided are an apparatus, a method, and a computer-readable medium, the apparatus including: an acquisition unit that acquires measurement data indicating a state of an object; and a supply unit that supplies the measurement data acquired by the acquisition unit to a plurality of classification models that complete learning by learning data for each of different periods including the measurement data for a period in which the state of the object is normal. And classifies the measurement data as either normal or abnormal according to the input of the measurement data; and a determination unit that determines, on the basis of a plurality of classification results output from the plurality of classification models, whether the state of the object is normal or abnormal.

The invention relates to a method and a device for generating a report under the assistance of a label, and the method comprises the following steps: 1) extracting a structured label set from a text report of a sample based on a large language model, wherein the structured label set comprises a multi-classification group consisting of dichotomous labels and mutual exclusion options; 2) aggregating the labels in batches, after a threshold value is reached, merging and de-weighting, performing specification and mutual exclusion group merging on synonymous, near-synonymous and redundant labels, and converging into a unified label library; 3) based on the text report and the tag library, outputting a tag subset of each sample through a large language model; 4) multi-modal multi-label classification model training: extracting each visual modal feature, performing weighted aggregation and splicing, and outputting each label group logits through a classification head to perform weighted group loss optimization; (5) carrying out joint training on the multi-modal large language model by using samples of'only images-reports' and'images + labels-reports', and (6) carrying out label prediction and screening on the images by using the classification model, and inputting'images + prediction labels' into the multi-modal large language model to obtain a final report.

The present invention discloses a test-time adaptation method for a calibration-free brain-computer interface, the method comprising the following steps: obtaining labeled training data from multiple source domain users to establish classification models for multiple categories; establishing the classification model, performing Euclidean alignment on the training data of each user, then merging all source domains, training multiple classification models using traditional cross-entropy loss, and obtaining multiple trained source domain models; obtaining the next test sample on the target domain data stream of the current user, performing incremental Euclidean alignment, and inputting the data into the trained multiple source domain models to obtain predicted probability values based on the multiple source domain models on the target domain data; using a spectral meta-learner method to perform integrated prediction of multiple models to obtain the predicted value of the current test sample. Based on the predicted probability value, each model is optimized separately by minimizing conditional entropy and using adaptive marginal distribution constraints within the batch to obtain optimized models adapted to the target domain; the present invention takes into account the real-time application of the cross-user brain-computer interface system, and adaptively adjusts the model without adding a calibration link.

The application provides a model training method, an audio classification method, a device, a medium and a program product, and mainly relates to machine learning technology in the field of artificial intelligence. The training method comprises the following steps: obtaining a first audio, actual classification results and actual position encoding results of the first audio in multiple classification dimensions; inputting the first audio into a target neural network model to obtain predicted classification results and predicted position encoding results of the first audio in the multiple classification dimensions; obtaining a classification loss according to the actual classification results and the predicted classification results; fusing the actual position encoding results of the first audio in the multiple classification dimensions to obtain actual fusion results, and fusing the predicted position encoding results of the first audio in the multiple classification dimensions to obtain predicted fusion results; obtaining a position encoding loss according to the actual fusion results and the predicted fusion results; and training the target neural network model according to the classification loss and the position encoding loss, so that the classification accuracy can be improved.

The present disclosure relates to the field of energy technology, and provides a carbon emission measurement method and device based on a smart park operation and management scenario. The method includes: determining the scenario category of the operation and management scenario; collecting a feature data set related to the carbon emission behavior corresponding to the scenario category; classifying the feature data set to obtain multiple classification data subsets, each classification data subset including at least one classification data; assigning a carbon emission conversion coefficient to each classification data in each classification data subset; if it is determined that a preset carbon emission measurement model library contains a target carbon emission measurement model that matches the carbon emission behavior category and carbon emission conversion coefficient of multiple classification data subsets, then the target carbon emission measurement model is retrieved; according to the target carbon emission measurement model, the total carbon emissions under the operation and management scenario are calculated, which can effectively and flexibly calculate carbon emission measurement under different daily management and operation scenarios.

Embodiments of the present disclosure disclose a sea surface oil spill detection method based on decision fusion of multi-core classification models. The method includes: obtaining a multi-spectral remote sensing image of a target sea area as an image to be detected; extracting features from the image to be detected to obtain feature information; constructing input data based on the feature information; inputting the input data into classification models of multiple single-core functions respectively to obtain multiple classification results; and obtaining a decision fusion result based on the multiple classification results to distinguish the oil spill area in the image to be detected. The sea surface oil spill detection method provided by the present invention inputs the input data into classification models of multiple single-core functions respectively to obtain multiple classification results, and then performs decision fusion on the multiple classification results to obtain a decision fusion result. The multi-core decision fusion algorithm can complement the advantages of each single core, improve the classification accuracy, effectively detect oil spills under a complex water color background, and further improve the accuracy of sea surface oil spill detection.

The application is suitable for the technical field of data processing, and provides a software type detection method and device, terminal equipment and storage medium, the method comprises the following steps: obtaining software data of a to-be-detected software; extracting software features of the to-be-detected software from the software data, wherein the software features comprise operation code features, text features, permission features and image features; classifying the to-be-detected software according to the software features to obtain multiple classification results of the to-be-detected software; determining composite software features of the to-be-detected software according to the multiple classification results; and determining the software type of the to-be-detected software according to the composite software features. The application can solve the defect that the use of a single data type leads to low detection accuracy in the related art, help to improve the accuracy and robustness of detection, and does not need to update a feature library, thus helping to reduce the maintenance cost.

This invention discloses a product classification and recognition method and system based on AI technology, belonging to the field of AI recognition technology. By constructing a semantic alignment master vector Edis, the system achieves structured extraction and ambiguity resolution of the overall semantic representation of the product. On this basis, the system introduces the construction of a product classification label semantic vector set Gvec and a semantic similarity score set Sset, enabling the system to perform fine-grained matching of multiple classification labels based on contextual semantic features. This constructs a confidence score Conf reflecting the classification credibility, which is then compared with a preset classification judgment threshold Thrs to ensure that product labels are output only when the classification judgment credibility is sufficient. Through the connection of the above processes and the linkage of output results, this method not only significantly reduces the probability of misjudgment caused by expression ambiguity and word order reversal, but also provides a reliable rejection mechanism when accurate classification is not possible, effectively improving the overall robustness and practical value of the classification system.

The invention relates to a fund multi-dimensional quantitative ranking method and system, and the method comprises the steps: collecting the real-time position data of a target fund, and dividing the fund types based on all ranking framework dimensions; calculating each quantitative index value of the target fund in each time window, and obtaining the similar ranking, ranking level and similar mean value of the target fund according to the quantitative index values; obtaining structured input data of the target fund based on the structured cue word, and submitting the structured input data to the large language model to generate a natural language report, the natural language report including fund analysis, risk cue and adaptation suggestion; and repeating the steps based on each ranking framework dimension, generating a natural language report of the same target fund under different fund categories, and constructing a multi-dimensional ranking system of the target fund. A fund multi-dimensional quantitative ranking system is constructed by covering a plurality of classification dimensions, a plurality of time periods and a plurality of indexes and giving appropriate comments to each index data.

The application discloses a kind of material handling control method and system with classified goods allocation function, it is related to the process control technical field based on big data analysis, based on high priority attribute subset, fusion path distance calculation generates optimization path sequence;On the basis of extracting path node conflict, temperature cross influence and time window overlap and other mutual restraint relations, iteratively adjust storage occupancy and handling order, get balanced order list;Finally, combined with material scheduling demand parameters and cargo classification dimension determines target handling instruction set, effectively eliminates the core contradiction of emergency material delay, temperature-sensitive material deterioration and low handling efficiency, realizes when receiving a batch of goods allocation instruction containing multiple attribute constraints, according to current specific circumstances automatically determine the relative importance of each dimension, and generate a set of reasonable handling instruction that meets multiple classification requirements simultaneously.

The invention relates to the technical field of intelligent classification garbage cans, in particular to an intelligent classification garbage can. Comprising a garbage can main frame, a plurality of classification boxes arranged in the garbage can main frame, a garbage temporary storage module, an identification module, a mechanical pick-and-place module and a visual control module for controlling the garbage temporary storage module and the mechanical pick-and-place module based on the identification module. According to the method, the classification accuracy can be improved, different types of garbage can be accurately distinguished, the garbage classification accuracy can be improved, the garbage can be quickly classified and treated, compared with manual classification, the treatment speed is higher, a large amount of garbage can be treated in a short time, various resources can be better recycled through accurate classification, and the method is suitable for popularization and application. The recycling rate of resources is improved, various valuable resources are contained in different types of garbage, the resources can be effectively separated out and reprocessed and reused, resource waste is reduced, and harmful garbage can be accurately separated from other garbage.

The present invention relates to a pattern recognition-based engine fault warning method, device, and storage medium. The method comprises: obtaining and preprocessing raw engine data; performing feature selection on the data using a random forest model as training data; selecting multiple classification models and connecting the selected classification models based on weights to form a fault recognition model, and using the training data to perform hyperparameter tuning; performing a performance evaluation on the trained fault recognition model; reselecting a classification model or re-adjusting hyperparameters if the model does not meet the requirements; and executing the next step if the model meets the requirements; establishing an engine fault warning model based on the trained fault recognition model; and establishing model evaluation indicators to evaluate the performance of the engine fault warning model; adjusting the hyperparameters of the engine fault warning model if the model does not meet the requirements; and performing a fault warning using the trained engine fault warning model if the model meets the requirements. Compared with existing technologies, the present invention has the advantages of high prediction accuracy.

The invention discloses a hyperspectral dry pea thermal damage identification method fused with multilayer wavelet decomposition characteristics. According to the method, spectral information of dry peas within the range of 400-1000 nm is obtained through a hyperspectral imaging technology, seven-layer decomposition is carried out on an original spectrum by adopting a db6 wavelet function, and multi-scale detail features are extracted. And an optimal wavelet layer is screened in combination with correlation analysis, and a multi-scale fusion feature set is constructed, so that the discrimination capability of the model on the heat loss grade is improved. And carrying out modeling analysis on the features by adopting a plurality of classification models such as PLS, SVM, LR and RF, and comparing the recognition performance of the features under different feature combinations. According to the method, redundant information interference is effectively reduced, the expression ability of the spectrum to the thermal damage characteristics is enhanced, and the accuracy and stability of the model are remarkably improved. The method has the advantages of non-contact, high throughput and intelligence, and is suitable for heat damage identification and quality grading of agricultural products such as dry peas.

The application relates to a method and system for mining and analyzing the correlation of clinical comorbidity of discharged patients. The method comprises the following steps: constructing an individual diagnosis and treatment narrative graph for each patient according to the discharge medical record text, generating a time sequence transaction sequence corresponding to each patient, and constructing a sequence database; based on the database, the original time sequence frequent pattern set is constructed by analyzing through an improved generalized sequence pattern algorithm; each pattern in the set is classified to obtain multiple classification clusters, and the original time sequence frequent pattern in each classification cluster is processed through multi-sequence alignment to construct a generalized clinical path graph, and the information in the generalized clinical path graph is extracted to generate a natural language abstract. The method improves the time sequence logic, knowledge abstraction degree and clinical interpretability of the clinical comorbidity correlation mining by constructing a diagnosis and treatment narrative graph, mining a time sequence frequent pattern, and constructing a generalized clinical path graph through clustering and multi-sequence alignment.

The application provides a ship type classification prediction method and system based on K-means and XG-Boost, acquires ship data and carries out pretreatment, then adopts a K-means clustering algorithm to respectively cluster each kind of data in the pretreated ship data to obtain multiple clusters, calculates error sum of squares of all data in each cluster, calculates the classification number of each kind of data according to the error sum of squares, selects a certain classification number by using an elbow method, marks the ship type of all clustered ships according to the classification number, then takes the ships with marked ship types as training set samples, adopts an XG-Boost classification algorithm to train the training set samples to obtain multiple classification prediction models, verifies the multiple classification prediction models to obtain an optimal classification prediction model, and predicts the ship types of all ships in the world according to the optimal classification prediction model and marks the ship types. The application can accurately classify all ships in the world, and can avoid overfitting and underfitting of the model while ensuring the accuracy.

The invention provides a wind power plant output active power control method and system, and the method comprises the steps: obtaining target energy values issued by a power grid at a plurality of time points in a preset time period and output active power values of a plurality of wind power plants at the plurality of time points in the preset time period; obtaining a frequency adjustment demand feature vector according to target energy values issued by a power grid at multiple time points in a preset time period and a first multi-scale neighborhood feature extraction module; obtaining a plurality of active power feature vectors corresponding to the plurality of wind power plants according to the plurality of output active power values of the plurality of wind power plants at the plurality of time points in the preset time period and a second multi-scale neighborhood feature extraction module; according to the frequency adjustment demand feature vector and the plurality of active power feature vectors, obtaining a plurality of classification feature matrixes corresponding to the plurality of wind power plants; inputting the plurality of classification feature matrixes into a pre-trained classifier model to obtain a plurality of classification results; and controlling the output active power of the corresponding wind power plant according to the classification result.

The utility model provides a marine organism collecting and classifying device, and mainly relates to the field of organism collecting and classifying equipment. A marine organism collecting and classifying device comprises a collecting barrel. The fish collecting device has the advantages that fishes of different sizes are classified according to the sizes by arranging the collecting barrels separated by the multiple stages of classifying plates, the rotating structures and the baffles are further arranged so that all the partitions can rotate, and influences between the fishes of different sizes are prevented through separation of the baffles; furthermore, a water inlet and outlet structure is arranged in the device, so that seawater flow can be controlled from the upper part to flow into the classification device to provide a living environment for marine fishes and discharge the internal water flow, and a multi-stage classification plate is arranged to be of a spiral structure, so that the multi-stage classification plate can be matched with a baffle to rotate to drive the fishes to be moved out step by step in a multi-stage manner; mutual influence caused by simultaneous discharge of fishes with different sizes is avoided.

The invention discloses a design scheme recommendation method and device, a storage medium and a computer program product, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining a plurality of advertisement pictures clicked at high frequency, inputting each advertisement picture into a pre-trained multi-label extraction model, obtaining a plurality of classification labels of each advertisement picture under different classification dimensions; label co-occurrence analysis is carried out on each classification label to obtain multi-label co-occurrence information, the multi-label co-occurrence information comprises at least one label matching scheme, and each label matching scheme comprises at least two classification labels; and if a design recommendation request input by a user is received, determining a target classification tag corresponding to the design recommendation request, generating a target design scheme according to the target classification tag and the multi-tag co-occurrence information, and taking the target design scheme as a recommendation result. The design efficiency of the advertisement picture is improved, and the design threshold of the advertisement picture is reduced.

The landslide analysis method based on the sensitivity of risk factors solves the problem of the existing difficulty in landslide prevention analysis and belongs to the field of disaster prevention technology. The present invention includes: collecting data of landslide disaster points and various landslide risk factors; dividing classification feature intervals for the raster data of each landslide risk factor to obtain a variety of different classification schemes, and each classification scheme includes multiple classification feature intervals; constructing a classification model for each landslide risk factor by using the deviation sum index and the fitting deviation sum index, and selecting the classification scheme when the classification model value is closest to 1; optimizing the classification feature intervals of the selected classification scheme; calculating the sensitivity DC of the landslide risk factor in the optimized classification feature interval; if DC>0, it means that the landslide is likely to occur in the corresponding classification feature interval; if DC<0, it means that the landslide is not likely to occur in the corresponding classification feature interval; if DC≈0, it means that it cannot be determined whether the landslide is likely to occur in the corresponding classification feature interval.