432 results about "Binary classification" patented technology

The invention discloses an emotion recognition method and system based on cross-modal alignment and matching enhancement. According to the method, firstly, feature extraction is carried out on text, audio and video modalities in a data set, and then a text and audio cross-modal emotion alignment module and a text and video cross-modal emotion alignment module are constructed respectively, so that cross-modal semantic alignment is realized. Constructing an emotion label matching module based on an alignment result, generating modal pairs with similar emotions but different labels by using a difficult negative sample mining strategy, and paying attention to cross-modal emotion consistency through a dichotomy task guide model; performing modal feature fusion on the three modals through a six-layer attention crossing mechanism, finally splicing feature vectors, inputting the spliced feature vectors into a long-sequence context fusion modeling module for deep modal fusion, and capturing cross-modal interaction information; and the fused features are sent to an emotion classification module, and a final emotion category recognition result is output.

The invention provides a multi-receptive-field network abnormal flow detection method and system based on endogenous security attributes, and relates to the technical field of network security and abnormal flow intelligent detection. The method comprises the following steps: firstly, performing multi-scale flow representation, preprocessing and data enhancement on network flow data to obtain enhanced input flow data; local features are extracted through basic convolution, and local and global fusion features are obtained based on a double-branch network comprising a multi-receptive field convolution branch and a Mama-self-attention branch; deep fusion representation is formed through multi-round feature extraction and tensor fusion, and finally binary classification and fine-grained classification results are output through global pooling and a linear classification layer. According to the method, high-precision, high-robustness and high-real-time detection of the abnormal traffic of the complex network is realized under low calculation overhead.

A system and method for facilitating training of large language model based recommender systems are provided. The system may utilize one or more LLMs to create probability distributions for binary classification tasks associated with specific user-item pairs. The probabilities may be utilized to rank one or more tasks directly. The training of the one or more LLMs may involve the use of Knowledge Distillation methods and may be based on incorporating a dual-label system such as, for example, hard labels and soft labels. The one or more LLMs training data may consist of user-item pairs and their corresponding features. The labels used in the training process may include binary classification labels and their respective probabilities. The system may further implement the trained one or more LLMs to determine rankings or recommendations associated with user engagement of one or more content items.

The invention belongs to the technical field of medical image intelligent diagnosis, and provides a breast cancer focus benign and malignant discrimination method based on a gated multi-expert mechanism. The method comprises the following steps: firstly, carrying out standardization and semantic preprocessing on a mammary gland X-ray image, a BI-RADS imaging report and structured clinical data, embedding age, mammary gland density and focus position information into a text template in a natural language form, and realizing unified expression of multi-modal input; secondly, extracting image features by utilizing a ResNet network and a simplified CLIP model, obtaining a text semantic vector by adopting a Bio-ClinicalBERT model, and establishing two sub-paths of a lump expert and a calcification expert in a Transform structure; further, an expert weight is dynamically generated through a gating routing mechanism, and soft routing fusion is executed; and finally, outputting benign and malignant results of the breast cancer focus by the binary classification module. According to the method, deep fusion and dynamic collaboration of the mammary gland X-ray image, the BI-RADS text and the clinical information are realized, and the accuracy and interpretability of breast cancer discrimination can be remarkably improved.

A bot detection service associated with an overlay network operates to score traffic as a probability of being a bot, as opposed to returning a binary classification (i.e., bot or human). According to the approach herein, scoring is determined through probability estimates, wherein a score (the probability) is based on considering a set of detections concurrently. In one embodiment, all (or substantially all) triggered (current) threat detections contribute to the score. The preferred approach penalizes requests that fail all (or substantially all) combinations of detection algorithms. According to a further feature, an automated tuning (autotuning) is also applied, e.g., using real-time empirical statistical models, to adapt the measurement of false positive probability for one or more threat detection algorithms to suit customer traffic trends. The approach herein is also extensible to include any number of future threat detection algorithms.

The invention relates to a Parkinson's disease walking state recognition method based on tensor singular value decomposition and automatic hyper-parameter optimization. The Parkinson's disease walking state recognition method comprises the following steps: acquiring a sensor signal of a target person; constructing a tensor according to the sensor signal; performing t-SVD decomposition processing on the tensor to obtain a frequency domain core tensor; the frequency domain core tensor and the sensor signal are subjected to feature extraction, the extracted features are input into a trained classification model, the binary classification recognition result of the Parkinson's disease patient and the healthy person is obtained, and the trained classification model is obtained through training of a training set marked with a patient and health contrast label. According to the method, efficient dimension reduction and noise suppression are realized by using t-SVD, and adaptive optimization is performed on the key hyper-parameters of the classification model in combination with an automatic machine learning technology, so that the dichotomy recognition accuracy and robustness of the Parkinson's disease patient and the health control are improved.

The invention provides an attack technology and tactical identification method oriented to honey array linkage. The method comprises the following steps: coding according to malicious sample data and associated TTP labels, and constructing a training data set; constructing an interpretable discrimination model for binary classification modeling to obtain a behavior triggering weight vector and an offset item of the malicious sample; extracting dominant trigger rules to construct a mapping inference rule base; obtaining alarm data, extracting a behavior clue based on a preset standard behavior dimension set, converting the behavior clue to obtain a trigger identifier, and mapping the alarm data into an alarm behavior vector; constructing a prediction TTP label set of the alarm behavior vector and recording a trigger path; and encapsulating the behavior semantic data, constructing an execution strategy response rule set, performing mapping in the execution strategy response rule set according to the behavior semantic data to obtain an execution strategy, and adjusting honey array deployment according to the execution strategy. By applying the method, accurate identification and transparent reasoning of the potential TTP in the attack activity can be constructed and realized, and linkage scheduling of a defense system is supported.

The present invention relates to the technical field of computer vision, and provides a pseudo-label filtering-based online domain change continual learning method and system. The method comprises: acquiring a pre-trained model, using the pre-trained model to predict changing target-domain data, and generating a pseudo-label for online adaptation; deriving a lemma for threshold-based pseudo-label filtering in online domain change continual learning on the basis of binary classification, and designing a threshold setting principle in the online domain change continual learning on the basis of the lemma; using the designed threshold setting principle to filter a pseudo-label having a low confidence level in model prediction, and introducing a class prior alignment method to encourage the model to perform fair prediction on an unknown-domain sample; and using the filtered pseudo-label to update and optimize the model to obtain a classification prediction result in the online domain change continual learning. In the present invention, an adaptive threshold capable of adapting to a CTTA process is established, thereby ensuring the quality of pseudo-labels.

The invention discloses an electric vehicle rear-end collision accident prediction method and system based on large model knowledge and trajectory data. According to the method, two vehicle types of an electric vehicle and a fuel vehicle are identified based on license plate colors, the vehicle types and vehicle tracks are bound, and the following process is divided into three states according to the combination of the front and rear vehicle types. On the basis, multi-dimensional features extracted from historical tracks are input into a large language model, expert-level rule knowledge used for distinguishing accidents and non-accidents is extracted through structured cues, and a final feature matrix is constructed according to the knowledge. And finally, based on the final feature matrix, performing differentiation processing on the three car-following states, training and calibrating a dichotomy model, and realizing real-time estimation of the rear-end collision probability at a specific moment in the future. According to the invention, by fusing the reasoning ability of the large model and the real trajectory data, the accuracy and interpretability of rear-end collision risk identification under the electric vehicle participation scene can be significantly improved, and the method has engineering deployment value.

This invention relates to the field of medical testing, specifically to a set of biomarkers, reagent kits, and their applications for diagnosing hypertension in children. This invention involves collecting tongue / intestinal samples from obese children with hypertension, obese children, and healthy individuals, performing metagenomic sequencing, and statistically analyzing the sequencing data using bioinformatics to identify disease-related tongue / intestinal flora. By integrating tongue / intestinal flora with disease information, a combination of flora biomarkers is obtained. A binary classification prediction model constructed using this combination can maximally detect hypertension in obese children.

The invention relates to the technical field of crossing of remote sensing image processing and computer vision, and discloses a remote sensing image change detection method which comprises the following steps: obtaining a first time phase bottom layer pixel feature and a second time phase bottom layer pixel feature according to a first time phase remote sensing image and a second time phase remote sensing image; according to the first time phase bottom layer pixel features and the second time phase bottom layer pixel features, enhanced visual features are obtained; according to the first pixel-text similarity score plot and the second pixel-text similarity score plot, semantic guidance features are obtained; according to the enhanced visual features and semantic guidance features, obtaining fusion features of multiple levels; pixel-level binary classification is carried out on the fusion features of the multiple levels, and a change detection binary image is output. According to the invention, the precision and accuracy of remote sensing image change detection are improved.

The invention discloses a power frequency waveform classification method and device, a medium and equipment, and belongs to the field of waveform classification, and the method comprises the steps: obtaining preprocessed power frequency waveform data of a power transmission line, calculating various waveform features, and inputting the features into a preset cascade classifier model. The model is composed of a plurality of XGBoost binary classifiers which are cascaded in sequence, and each classifier is responsible for executing a specific binary classification task. The output result of the previous classifier is used for triggering the starting of the next classifier, or the final classification result is directly output. The cascade structure not only improves the accuracy and efficiency of classification, but also reduces the complexity of the model, and effectively solves the problem that the fault positioning power frequency of the power transmission line cannot be accurately and efficiently classified in the prior art.

The invention discloses a bidirectional information retrieval enhancement generation method for a large language model, and belongs to the technical field of artificial intelligence. In order to overcome the defects that noise is introduced and key evidences are omitted due to the fact that traditional RAG only executes'query-document 'one-way retrieval, a two-way semantic perception retrieval enhancement generation model and a two-stage training framework are constructed, wherein in the first stage, the positive / negative example distance is increased in an embedded space in a contrast learning self-supervision mode; in the second stage, fine-grained correlation discrimination is carried out on query-document bidirectional sentences through supervised dichotomy, and probabilistic correlation scores are output; in the reasoning stage, the bidirectional probabilities are fused according to Bayesian to obtain final relevancy, document reordering is carried out, and plug and play can be achieved without fine adjustment of LLM in the whole process. According to the method, the accuracy and consistency of single-hop and multi-hop questions and answers and fact checking tasks are remarkably improved, and the method has the advantages of light weight and low deployment cost.

The invention discloses an illusion detection method and device for a large language model and a medium, and relates to the technical field of natural language processing and artificial intelligence. The method comprises the steps of extracting internal state data when an answer is generated by a large language model; wherein the internal state data comprises a hidden state vector of the last token when the answer is generated and the number of tokens for generating the answer; performing feature engineering processing on the internal state data to obtain an internal confidence score; judging through an external language model to obtain an external judgment score; fusing the internal confidence score with the external judgment score to obtain a comprehensive confidence score; and outputting an illusion or non-illusion binary classification result according to whether the comprehensive confidence score exceeds a predetermined threshold.

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 invention discloses a depressive emotion intelligent identification and intervention system. The system comprises a data acquisition module used for acquiring original data of a user in real time; the data preprocessing module is used for performing layered preprocessing on original data; the feature extraction module is used for extracting each modal feature; the multi-modal fusion module realizes multi-modal dynamic fusion based on a bidirectional long short-term memory network and a cross attention mechanism, and supports multi-task output of depression dichotomy, PHQ-9 scale regression prediction and the like through a full-connection network; the user interaction module is used for providing task guidance, data acquisition control and intervention suggestion visualization for a user; the system integrates a self-developed VR (Virtual Reality) mind game and an AI (Artificial Intelligence) psychological big model, and provides intervention tools such as mind training, dialogue intervention and PC (Personal Computer) terminal severe games for different depression levels. The data management module is used for storing user information, original data, a depression emotion recognition process and evaluation result data, and a feasible technical means is provided for early recognition and auxiliary evaluation of depression risks.

The invention discloses a policy document identification and judgment method and system based on a large model, and the method comprises the following steps: firstly, carrying out the preprocessing of an original document, completing the format conversion and text cleaning, and extracting at least one basic feature; inputting the preprocessed document into a pre-training dichotomy model, and obtaining a probability value P1 belonging to the policy document; screening P1 through a first threshold value T1 and a second threshold value T2, clearly judging a policy or non-policy document, and if a fuzzy interval is processed, executing keyword matching analysis and calculating a keyword matching score S; and screening S through a third threshold T3 and a fourth threshold T4, when S is still a fuzzy interval, extracting multi-dimensional policy characteristics to construct a structured context, inputting the structured context and the original text of the document into a large language model, reasoning according to a preset cue word template, and outputting a result containing confidence and a judgment reason. The policy document identification method has high accuracy, strong interpretability and good adaptability, and can efficiently, accurately and transparently identify the policy document.

The invention discloses a radar target detection method based on graph node dual-channel feature attention fusion, and belongs to the technical field of radar signal detection, and the method comprises the following steps: 1, carrying out the graph node division of received frame radar echo data; 2, respectively extracting time domain amplitude and time frequency characteristics from echo time sequence data corresponding to each graph node; step 3, establishing a feature preprocessing sub-network; step 4, constructing a node feature fusion sub-network; 5, constructing a signal classification graph neural sub-network; 6, connecting the feature preprocessing sub-network, the node feature fusion sub-network and the signal classification graph neural sub-network in series to form a radar target detection neural network; and 7, inputting the test set into the trained radar target detection neural network, and outputting a dichotomy result of which the corresponding node is a target or clutter signal. Through the scheme, the target detection capability of the radar in the clutter environment can be improved.

The invention discloses a lithofacies paleogeography single factor analysis method, medium, equipment and product, and relates to the technical field of lithofacies paleogeography analys.The method comprises the steps that lithofacies description, well position coordinates and stratigraphic stratification data in logging interpretation data are obtained; performing binary classification on lithofacies types according to whether the lithofacies description accords with the definition of the target lithofacies; based on the stratigraphic data and the binary lithofacies type, three single factor indexes of each well are calculated, and the three single factor indexes are the total stratum thickness, the target lithofacies cumulative thickness and the rock-to-ground ratio; according to the well position coordinate data and the three single-factor indexes, well points in the whole area are paired in pairs, experimental variation function values under different lag distances are calculated, and experimental variation scatter diagrams of the three single-factor indexes are drawn respectively; kriging gridding interpolation is carried out on the scatter points, contour maps of the three single-factor indexes are drawn, and lithofacies paleogeography analysis is carried out by using the three contour maps. The lithofacies paleogeography single factor analysis is realized based on the logging interpretation data.

The invention discloses an intelligent detection method and system for AI texts, and relates to the technical field of AI text detection.The method comprises the steps that AI texts and non-AI texts corresponding to multiple themes are obtained, and a training set of theme contrast is constructed; performing word segmentation processing on each text in the training set, and quantifying vocabulary features of each text; analyzing each sentence in each text in the training set, and quantifying syntactic features of each text; on the basis of the vocabulary features and syntactic features of all the texts, the style feature distinction degree of each theme is calculated; calculating the text weight of each subject based on the style feature distinction degree, constructing a dichotomy model, and improving a loss function of the model in combination with the text weight; and training the model based on the improved loss function and the training set, and realizing intelligent detection of the AI text through the trained dichotomy model. According to the method, excessive dependence of the model on specific content vocabularies can be effectively avoided, and adaptability and discrimination accuracy of the model to different topic texts are improved.

The invention relates to the technical field of vehicle networks and deep learning, in particular to a two-stage vehicle network intrusion detection method, which comprises the steps of inputting a to-be-detected vehicle data sequence into a trained intrusion detection model, and outputting a corresponding attack detection result; the intrusion detection model is processed by the following steps: S201, inputting a to-be-detected vehicle data sequence into a two-dimensional convolution classifier, and outputting a corresponding binary classification result; s202, if the dichotomy result is attack data, executing the step S203, and if the dichotomy result is legal data, executing the step S205; s203, inputting the data sequence of the vehicle to be detected into the Robust LSTM classifier, and outputting an attack type label; s204, outputting the corresponding attack data labels and attack type labels as attack detection results; and S205, outputting the corresponding legal data label as an attack detection result. According to the invention, the real-time performance, pertinence and reliability of Internet of Vehicles intrusion detection and an Internet of Vehicles system are improved.

The invention relates to a riverway ice condition identification method based on optics-SAR fusion and adaptive segmentation, and belongs to the technical field of remote sensing image processing and application. Riverway ice condition features are extracted from the optical remote sensing image, and a Ka-SAR feature map is extracted from the Ka-SAR image by adopting an improved high-resolution network; carrying out multi-modal and multi-scale feature fusion on the extracted features, carrying out scale specificity feature extraction by adopting a Gaussian pyramid, and then carrying out weighted fusion on the river ice condition features and the Ka-SAR features on each scale based on a scale specificity weight distribution principle; aggregating the multi-scale fusion features into a final fusion feature map by adopting a bottom-up pyramid reconstruction strategy; improved Kuan filtering is used to optimize the fused feature map, and adaptive threshold segmentation is used to realize ice surface and non-ice surface binary classification in the feature map. The method can achieve the precise segmentation of the ice condition region, and improves the recognition precision of the thin ice region.

The invention belongs to the technical field of forgery detection, and particularly relates to a deep forgery detection method based on a rank adaptation network, and the method comprises the steps: carrying out the initial feature extraction of an input image through a CLIP-ViT backbone network; constructing a dynamic rank adjustment module comprising a dynamic rank controller, a learnable basis matrix and a dynamic rank fusion unit, inputting the initial features into the dynamic rank adjustment module, adaptively adjusting the rank of the feature space through the dynamic rank controller, the learnable basis matrix and dynamic rank fusion, performing projection, and outputting rank enhancement features; the rank enhancement features are sent to a multi-scale attention module, multi-granularity counterfeit traces are aggregated, and refined multi-scale fusion features are output; and inputting the multi-scale fusion features into a classifier to complete binary classification decision of image authenticity. According to the method, on the basis of maintaining the detection competitiveness in the domain, the generalization ability on an unknown data set can be effectively improved, and certain anti-interference robustness is considered.

The invention provides a video coding method, which comprises the following steps of: inputting a coding tree unit to be coded, and extracting texture features of a coding block, the texture features comprising local binary pattern similarity of all pixel points in the coding block and direction dispersion of each pixel point; inputting the texture features into the trained multi-type division prediction architecture, and sequentially judging whether to execute quadtree division, horizontal or vertical division, horizontal binary tree or ternary tree division and vertical binary tree or ternary tree division or not through a binary classification problem method; and calculating the rate distortion cost under each division mode to obtain an optimal division mode and a suboptimal division mode of the coding block, and outputting an optimal mode or a combination of the optimal mode and the suboptimal mode. According to the invention, the accuracy of the prediction mode in coding is improved.

The present application relates to the technical field of recipe recommendation, and particularly relates to a cooking auxiliary method and device based on recommended recipes. The method comprises the following steps: in response to an operation instruction of a user for a recommended recipe, identifying the operation instruction to obtain instruction information, and inputting the instruction information into a pre-trained binary classification model to enable the binary classification model to judge the type of the instruction information; in the case that the instruction information belongs to inquiry information, inputting the instruction information into a pre-trained recipe question and answer language model to enable the recipe question and answer language model to generate a response matched with the instruction information; in the case that the instruction information belongs to correction information, inputting the instruction information into a pre-trained recipe correction model to enable the recipe correction model to determine a step to be corrected in the recommended recipe to which the instruction information is directed, and adjusting the step to be corrected according to correction information contained in the instruction information; and providing the generated response or the adjusted recommended recipe to the user.

The invention relates to a fundus image quality evaluation method and system, an intelligent terminal and a storage medium, and relates to the technical field of fundus image quality evaluation, and the method comprises the steps: obtaining a to-be-evaluated fundus image; inputting the eye fundus image into a backbone network based on a ResNet18 network structure to obtain a multi-scale feature map; performing spatial size scaling and alignment processing on the multi-scale feature map to obtain a scaled and aligned feature map; carrying out series fusion on the scaled and aligned feature maps to obtain a fused feature map; adding the position coding matrix and the fusion feature map element by element to obtain a coding feature map; inputting the coding feature map into a multi-size feature fusion module to obtain a multi-size fusion feature map; inputting the multi-size fusion feature map into a plurality of feature optimization modules to obtain an optimized feature map; and splicing and fusing the plurality of optimized feature maps, and inputting the fused optimized feature maps into a full connection layer for binary classification so as to output a quality evaluation result of the eye fundus image. The method has the effect of improving the accuracy of eye fundus image quality evaluation.

The present disclosure relates to a method and device for extracting and detecting time-frequency joint features of epilepsy based on covariance decomposition, an electronic device and a storage medium. The method comprises: performing data preprocessing on collected electroencephalogram (EEG) data to generate EEG data; calculating a covariance matrix, eigenvalue decomposition, feature extraction and stretching processing on the EEG data after decentralization to generate a time-domain feature vector; performing spectral feature extraction and frequency-domain covariance feature extraction respectively to generate a spectral feature vector and a frequency-domain covariance feature vector; generating time-frequency joint features based on a preset feature fusion strategy; and classifying and identifying the time-frequency joint features based on a preset binary classification method to complete epilepsy prediction. The present disclosure extracts and fuses time-frequency multi-scale feature information, which can effectively shorten the manual data labeling time of medical workers, improve the labeling efficiency of epilepsy seizure events, and provide a new approach and method for clinical application of epilepsy detection.

The invention discloses a MALDI-TOF mass spectrum microbiological identification method based on the distance between spectrums, and the method comprises the following steps: 1, spectrum data preprocessing: carrying out denoising and baseline correction processing on original mass spectrum data, extracting and merging corrected data peaks, and carrying out normalization processing to form a data list; step 2, carrying out average reference spectrum construction on the data based on spectrum spacing dimension reduction, and collecting a plurality of representative spectrums for each microorganism; pairing the same spectrum by using a peak value matching algorithm; calculating the average position and intensity of each group of matching peaks to form an average reference spectrum; constructing an average reference spectrum, then carrying out inter-spectrum distance calculation, obtaining peaks corresponding to the sample spectrum and the average reference spectrum, then calculating an inter-peak distance, and generating a distance vector as a low-dimensional representation of the microorganism sample spectrum; and step 3, microorganism classification: performing binary classification and multi-classification on the microorganism sample to complete microorganism identification.

The application discloses a picture recognition model training and picture recognition method, system, device and medium. The method comprises the following steps: obtaining a target picture, performing feature extraction on the target picture based on a feature extraction network to obtain target features of the target picture; judging a picture type of the target picture based on a binary classification network and the target features; and determining a processing strategy for the target features according to the picture type obtained by the judgment. That is, after the feature extraction on the target picture, the target features obtained by the extraction are classified, different processing strategies are adopted for different types of picture types, the flexibility of the face recognition model training is improved, and the performance of the face recognition is improved.

The invention discloses a test paper direction correction method, electronic equipment and a storage medium, and the method comprises the steps: obtaining the position information of a preset line number of a test paper picture, and judging whether the angle of the test paper picture reaches a preset angle based on the position information of the preset line number, the preset angle comprising 0 degree and 180 degrees; if the preset angle is not reached, the position information of the preset line number is rotated by 90 degrees, the position information of the preset line number after 90-degree rotation is sent to a preset dichotomy model, and the preset dichotomy model can carry out dichotomy on the two angles; and determining the rotation angle of the test paper picture according to the output of the preset dichotomy model and whether the position information of the preset line number is rotated by 90 degrees, wherein the preset dichotomy model is obtained by training based on a deep learning model. By converting the four classifications into the two classifications, the accuracy of test paper direction correction can be improved.