174 results about "Class imbalance" patented technology

The invention discloses an irregular small target identification method under a non-high-definition complex background image. A multi-scale feature pyramid is constructed through bidirectional feature fusion, so that the feature expression ability of a small target is enhanced; applying a space-channel attention module to adaptively highlight target features and suppress complex background interference; by introducing a composite loss function including class balance focus loss and enhanced bounding box regression loss, model training is optimized to deal with class imbalance and improve the positioning precision of an irregular target. A self-adaptive multi-scale detection head is adopted, and dynamic feature fusion and scale perception branches are utilized to realize accurate detection of targets with different sizes; according to the method, the problems of low recognition precision and poor adaptability caused by weak features, background interference and irregular shapes of irregular small targets in low-resolution and complex background images are effectively solved, and the monitoring performance in actual applications such as unmanned aerial vehicle aerial photography and remote monitoring is remarkably improved.

The invention discloses an equipment anomaly tracing method and system based on a digital twinborn and graph neural network, and belongs to the technical field of industrial intelligent operation and maintenance and fault diagnosis. The invention provides an innovative solution integrating digital twin high-fidelity simulation and a graph structure deep learning algorithm, aiming at the technical bottlenecks that the generalization performance of an existing data driving method is insufficient under the conditions of fault sample scarcity and category imbalance and the traceability accuracy of unknown and composite faults is poor. The method comprises the following steps: constructing a high-fidelity digital twin integrating multi-dimensional physical attributes and a system topology structure; based on a fault mode, influence and harmfulness analysis method system, constructing a fault mode library comprising a plurality of single fault modes and composite fault modes, and generating an enhanced training data set with accurate labels through an automatic fault injection mechanism; training a graph neural network model with a multi-level attention mechanism by using the data set so as to learn a propagation rule of a fault in a complex system topology; and finally deploying the model to carry out abnormity traceability analysis on real-time industrial Internet of Things monitoring data. According to the method, the fault diagnosis generalization ability and the positioning precision under the sample imbalance condition are remarkably improved.

The invention discloses an Internet of Vehicles CAN bus intrusion detection method based on noise perception active learning, and belongs to the technical field of Internet of Vehicles safety and machine learning. The invention aims to solve the technical problems of false label noise interference, high manual labeling cost, high attack missing report rate caused by class imbalance and the like. The core of the method is to execute a noise sensing mixed query strategy in an iterative loop: firstly, generating a pseudo tag through clustering and correcting by using an integrated noise detector; secondly, calculating uncertainty scores and noise probabilities of the samples, fusing the uncertainty scores and the noise probabilities to obtain a comprehensive score, and preferentially selecting the samples with high uncertainty and low noise probabilities; and then adaptively selecting a sampling strategy according to the model performance and applying category balance constraint. The query batch is used to iteratively update the model while dynamically adjusting the classification threshold to reduce the missing report rate. According to the method, the influence of pseudo label noise can be effectively suppressed, and the attack detection precision and generalization capability are remarkably improved with extremely low labeling cost.

The invention belongs to the technical field of essential gene prediction, and particularly relates to an essential gene prediction method based on DNA large model and time-frequency domain deep learning fusion, and the method comprises the steps: taking a domain DNA large model as a core representation layer, and obtaining special gene representation through cross-species corpus pre-training and task fine tuning; a T-Block and F-Block dual-channel time-frequency fusion structure is adopted, and the local dependence and long-range regulation relation of a gene sequence is synchronously captured by expanding DFT (Discrete Fourier Transform), complex value attention and iDFT (Initial Discrete Fourier Transform) conversion; designing an efficient modeling reasoning scheme of sliding window slices and gene-level aggregation aiming at an ultra-long sequence; in combination with class imbalance and a noise robust training strategy, cross-cell line / cross-platform transferable threshold output is realized through temperature scaling calibration, an uncertainty quantization and structured interface is matched, and drug target screening and experimental design decision are supported. The system supports the realization of multiple programming languages, and can complete low-delay end-to-end reasoning in a conventional hardware environment.

The invention discloses a drug relocation model construction method for simultaneously predicting drug-target interaction and drug-disease incidence relation, and belongs to the field of drug research and development, and the method comprises the following steps: integrating heterogeneous networks and attribute characteristics of drugs, targets and diseases, learning multi-relation node embedding by using RGCN, and constructing a drug relocation model for simultaneously predicting drug-target interaction and drug-disease incidence relation; a Gelato algorithm is combined to enhance a network structure, an auto-covariance is introduced to calculate a potential association score, and drug-target interaction and drug-disease association are synchronously predicted; the weighted cross entropy and N-pair loss joint optimization is adopted, unbiased training is realized, the problems of class imbalance and network sparseness are solved, and the model generalization ability and prediction precision are improved.

The invention discloses a deep learning-based disordered material identification method and device, electronic equipment and a program product. The method is realized based on a trained identification model, a C3k2-ASL module is introduced into a backbone network, ASL Block in the module can model long-range structure dependence in the width direction and the height direction at low calculation overhead, and the perception ability and identification robustness of the model to slender, dispersed and low-contrast disordered material features are enhanced. In order to improve the modeling capability of the model on the overall spatial distribution of the disordered materials, an MGCA module is introduced into the neck network, and the accurate recognition capability of the model on different types of disordered materials in a complex city scene can be improved by extracting multi-dimensional global context information and realizing fusion through a dynamic attention mechanism. Aiming at the problem of class imbalance, a loss function is improved based on a constraint logarithm reweighting modulation mechanism, so that the model is degraded into uniform weighting during data equalization and is still stably optimized during long-tail distribution, and the generalization performance and the training stability of the model are improved.

Application of multi-task learning technique(s) to machine logic (for example, software) used to detect financial transactions that are fraudulent or at least considered likely to be fraudulent. Some embodiments include adjustments and / or additions to conventional multi-task learning techniques in order to make the multi-task learning techniques more suitable for use in fraud detection software. One example of this is compensation for class imbalances that are to be expected as between the likely-fraud and not-likely-fraud classes of data sets (for example, training data sets, runtime data sets).

The invention discloses a self-adaptive optimization method and system for online decoding of a motor imagery brain-computer interface, and the method comprises the steps: carrying out the real-time self-adaption of an electroencephalogram data stream of a target user through a teacher-student model framework on the premise that the privacy protection of source domain training data does not need to be accessed; performing batch weight normalization during testing, decoupling normalization statistic updating and parameter optimization by stopping gradient operation, and stabilizing feature representation; a dynamic category specific entropy threshold mechanism is combined with online category frequency and batch confidence to adaptively screen a high-confidence sample for each category; a dynamic online reweighting strategy is designed, and weights are distributed according to the sample entropy and the category frequency to balance the optimization process; and decoupling contrast learning based on a fixed prototype is introduced, and feature space distribution is optimized. According to the method, the problems of statistic drift, poor fixed threshold adaptability, category imbalance sensitivity, insufficient feature optimization and the like are solved, and the adaptability, the stability and the robustness of cross-user motor imagery brain-computer interface online decoding are improved.

The invention provides a multi-mode command entity recognition method based on course comparative learning, and belongs to the technical field of natural language processing and computer vision. The method comprises the following steps: respectively carrying out semantic representation modeling on text and image data to obtain text features and image features; multi-level semantics from coarse granularity to fine granularity between the text features and the image features are aligned through staged comparative learning, and association between the text features and the image features is enhanced; fusing the text features and the image features through a gating multi-interest fusion mechanism to obtain fusion features; and performing sequence labeling on the fusion features by using a dynamically weighted conditional random field model. By combining a learnable gated multi-interest fusion module and a category-sensitive dynamic decoding mechanism, the problems of modal noise interference and category imbalance are effectively relieved, and the accuracy and robustness of multi-modal entity recognition and the applicability and stability of the model in a real social media environment are improved.

The invention provides a semantic segmentation method, system and equipment based on a traditional village multi-source image. The method comprises the following steps: acquiring a multi-source image and preprocessing the multi-source image; performing pixel-level labeling on the multi-source image to obtain a corresponding label graph; calculating the class imbalance weight of each class based on the tag graph, and obtaining a normalized class weight; inputting the label graph into a multi-scale context-aware coding-decoding network for feature extraction and fusion to obtain a prediction probability, constructing weighted multi-classification cross entropy loss based on a normalized category weight and the prediction probability, constructing a composite loss function in combination with boundary sensitive loss, and carrying out iterative updating on the network until convergence, so as to obtain a multi-scale context-aware coding-decoding network; through the setting, a synergistic effect is formed in four levels of data construction, a network structure, a training strategy and result output, so that objective indexes of traditional village street view scale semantic segmentation are improved, and the labor burden in an engineering use scene is remarkably reduced.

The application relates to a time series data feature enhancement representation method and device based on multi-scale contrast learning in the field of computer information technology. The method comprises the following steps: extracting three types of node relationship data and basic attribute information of each order node of original time series order data respectively, constructing an order node feature matrix, constructing a relationship graph for the three types of node relationship data by using a neighborhood subgraph order node feature aggregation unit, processing the order node basic features corresponding to each relationship graph, extracting global semantic features and local semantic features according to the order node basic features by using a global level order contrast learning unit and a local level order contrast learning unit respectively, and fusing the two features to obtain multi-scale fusion logistics order enhanced features after the order basic attribute is enhanced. By using the method, the sparsity and class imbalance of logistics order features can be relieved, multi-scale information fragmentation can be solved, and the order semantic expression and downstream task adaptability can be improved.

The present disclosure provides a system and a method (200) for detecting an agent vehicle entering a traffic lane. The method includes the use of a sensor for collecting (202) data from nearby vehicles and a control unit for extracting (204) features from this data. These features are then inputted (206) into a Long Short Memory (LSTM) model, which generates an n×1 output vector (208). A user-defined loss function is applied (210) to balance class imbalances in the output vector. The processor predicts (212) the probability of an entry maneuver at several future intervals based on the outputs of the LSTM model. If an entry maneuver is predicted, the processor adjusts the navigation or steering strategy of the agent vehicle accordingly (214).

The application discloses a kind of class imbalance table data processing methods based on contrast constraint diffusion generation, belong to data processing technical field, it includes using binary classification table dataset to generate model training, obtain the generation model based on noise generation minority class's synthetic latent representation, generate model is randomly generated with several noise inputs, diffusion modeling branch is denoised to each noise and obtains the synthetic latent representation of minority class of each noise, using decoding module is decoded to each synthetic latent representation, obtains synthetic table sample;Data filtering is carried out to all synthetic table samples, then all synthetic table samples after filtering are merged with table dataset, and enhanced imbalance table data are obtained.The application introduces contrast learning branch in the process of generating minority class sample, effectively overcome the deficiency existing in class imbalance table data processing of existing oversampling method and general generation model.

The application discloses a kind of rotary machinery class unbalanced multi-modal layered collaborative optimization fault diagnosis methods, comprising the following steps: S1: obtain the multi-modal monitoring signal of rotary machinery equipment under multiple different operating conditions S2: extract the high-dimensional feature representation of each mode;S3: execute category adaptive cross-modal comparison learning module, obtain stable feature representation with category differentiation;S4: execute two-stage attention multi-modal feature fusion, obtain cross-modal fusion feature;S5: construct category adaptive dynamic Gaussian kernel width, align cross-domain local distribution;S6: construct unified model training objective function, optimize fault diagnosis model.The application compared with prior art is in that: provide a kind of layered progressive collaborative optimization framework, realize multi-modal feature cross-modal stable alignment, two-stage purification fusion, category adaptive dynamic kernel cross-domain local distribution alignment and so on a kind of rotary machinery class unbalanced multi-modal layered collaborative optimization fault diagnosis method.

The invention discloses a class imbalance data model training method based on double views and semi-supervised learning, and the method comprises the following core steps: firstly, carrying out the data enhancement of an original image, and generating a weak enhancement view and two strong enhancement views; secondly, generating a depolarization false label with smaller deviation by utilizing a prediction result of the weak enhancement view and combining a dynamically updated global category confidence library; meanwhile, according to the original prediction of the amblyopia, active sampling is carried out to generate a pseudo-negative label for reverse supervision; and finally, through a multi-task optimization framework, combining depolarization loss, negative learning loss and consistency loss to jointly guide model training. According to the method, false label deviation is reduced through depolarization learning, the model discrimination capability is enhanced through negative learning, the model robustness is improved through consistency learning, and the learning effect and generalization capability of the model on class imbalance data can be remarkably improved under the condition of a small amount of labeled data.

The invention provides a dual confidence calibration method and system for a neural network classifier, equipment and a medium, and effectively solves the problems that an existing single-stage calibration method is difficult to give consideration to excessive confidence, under confidence, class imbalance sensitivity and the like. A mixed loss function fusing bifocus loss and difference between multi-class confidence and accuracy is introduced in a training stage, so that a neural network classifier is promoted to generate well-calibrated prediction distribution; a class-by-class multi-partition temperature scaling model optimized based on a coupling simulated annealing method is adopted in the reasoning stage, and calibration requirements of different classes and different confidence intervals are more accurately met compared with a traditional temperature scaling technology of a single temperature coefficient; in the training process, classifiers of different rounds are stored, multi-model calibration results are averaged class by class in the test stage, the calibration stability is effectively improved, prediction confidence errors are reduced, and the method is particularly suitable for the safety key fields such as medical diagnosis and automatic driving which have extremely high requirements for prediction reliability.

The application provides a spatial perception enhancement and pathological image classification method for processing class imbalance, aiming to enhance the model's perception of spatial information and process class imbalance. The specific steps are as follows: in the preprocessing stage, the image is segmented, the background is filtered out, and the image block is cut, and the coordinates are recorded; in the feature extraction and fusion stage, the pathological features are extracted through a pre-trained convolutional model, then the coordinates are normalized to form position features, and the pathological features and position features are fused to form comprehensive features with spatial information. In the reasoning stage, a multi-scale feature enhancement network is introduced, which can capture local features through convolution operations and model spatial relationships through attention mechanisms, thus achieving more comprehensive processing of local and spatial information. Finally, the parallel classifier is used to solve the problems of class imbalance and sample shortage, and the final pathological classification result is obtained. This method can better assist pathologists in improving the diagnosis accuracy and reducing subjective differences in the diagnosis of chondroma.

The invention belongs to the technical field of interpretable active learning, and particularly relates to a cell image attribute collaborative driven interpretable active learning method, which comprises the following steps of: in an inspection process, acquiring an image from a blood cell smear by using a camera, constructing a newly acquired cell image data set D1, randomly extracting a small amount of samples from the data set D1 to construct an initial labeling set D2; performing category labeling on the D2 and completing model pre-training based on the D2, performing standardization and preprocessing on the remaining unlabeled data in the D1, dividing the remaining unlabeled data into a plurality of batches according to the number to form a candidate unlabeled pool D3, sequentially inputting the data in the D3 into a model according to a preset batch, and for each batch of input data, adopting a class imbalance adaptive sample screening mechanism CASS to obtain a candidate unlabeled pool D3; the method comprises the following steps: comprehensively utilizing three types of indexes of uncertainty, representativeness and diversity under current weight configuration, introducing attribute information, carrying out weighted calculation in combination with a type frequency and an effective sample number, and screening to obtain candidate high-value samples of the batch;

The invention relates to the technical field of hierarchical text classification, and particularly provides a few-sample hierarchical text classification method based on a pre-training language model. The method comprises the following steps: generating a prompt of an adaptive task and a classification feature representation through a dynamic prompt generator; a semantic sharing vocabulary mapper is adopted, and semantic consistency is kept through coding father-son relations and cross-level label dependence; the dynamic weighting loss function is utilized, the loss contribution is adjusted according to sample difficulty, label hierarchy and category imbalance, the problem of hierarchical semantic confusion is solved, semantic sharing between labels is achieved, and dependence of a model on hierarchical labeling is reduced.

Class imbalance in a training dataset may negatively impact the accuracy of a machine-learning model in classifying rare events that are underrepresented in the training dataset. Training datasets comprising time-series data present a unique challenge. Accordingly, resampling techniques for up-sampling and / or down-sampling a training dataset of time series are disclosed. The up-sampling may respect the temporal correlation of time samples in the time series, while generating synthetic time series that mimic the feature values of time series belonging to the minority class. Down-sampling may be used to fine-tune the ratio of time series belonging to the minority class to the time series belonging to the majority class.

The invention discloses a femoral head necrosis prediction system and method based on Transform, and belongs to the technical field of image processing, the system comprises a data acquisition module, a data processing module, a data balancing module, a feature extraction module, a feature merging module and a necrosis classification module; the method comprises the following steps: acquiring preoperative hip joint CT data, pre-processing to unify specifications, adopting directional data enhancement to solve the problem of class imbalance, extracting features by virtue of Video Swin Transform, merging and modeling a spatial relationship through volume blocks, and finally outputting a necrosis classification result. According to the femoral head necrosis prediction system and method based on Transform provided by the invention, the prediction accuracy and robustness are effectively improved, the problem of model instability caused by strong subjectivity, hysteresis and data imbalance in the prior art is solved, an objective and efficient prediction tool is provided for clinic, and the risk of postoperative complications is reduced.

The invention relates to the technical field of machine learning. The invention discloses a class imbalance-oriented coevolution fair machine learning model integrated training method, which can balance the accuracy and fairness of model prediction. The method comprises the following steps: acquiring a training data set, wherein a worst verification subset of the training data set is a set subjected to fair violation calculation processing; performing training processing on the first target neural network set by adopting the training data set to obtain a second target neural network set; and screening the second target neural network set according to a preset rule to obtain a fair machine learning model, and applying the fair machine learning model to a personal credit approval application scene of a bank to predict a credit state of the user.

The invention discloses a few-sample sleep staging method and system based on dynamic weight, and the method comprises the steps: collecting original polysomnogram data, extracting an electroencephalogram signal channel, and carrying out the standardization processing; constructing a one-dimensional convolutional neural network as a feature extractor, and performing supervised learning pre-training; dividing a data set, and constructing a meta-learning task; inputting all samples in the support set and the query set into a pre-training feature extractor, and outputting corresponding support set sample feature vectors and query set sample feature vectors; calculating a feature prototype vector of each sleep stage, and calculating a dynamic weight of each stage; calculating the similarity between the feature vector of the query sample and the prototype vector of each stage, combining the similarity with the corresponding dynamic weight to obtain a weighted similarity score, and selecting the sleep stage with the highest score as a classification result; according to the method, the problems of poor model generalization and class imbalance are solved, and the technical performance is remarkably improved.

The invention relates to a cerebral apoplexy gait phase recognition method and system under class imbalance, and belongs to the technical field of cerebral apoplexy gait analysis. The method comprises the following steps: acquiring original gait signal data of a stroke patient; inputting the original gait signal data into a feature enhancement unit, and performing global average pooling, time sequence feature enhancement branch and channel feature enhancement branch processing to obtain a first fusion feature; inputting the first fusion feature into a long-short sequence feature extraction unit to obtain a second fusion feature; inputting the second fusion feature into a feature calibration unit to obtain a calibration feature; the calibration features pass through a linear layer classifier to obtain a stroke gait phase recognition result; in the training process, a self-adjusting cost-sensitive loss function is adopted to carry out parameter optimization. The accuracy of stroke gait phase recognition can be improved.

A SAR image change detection method based on a dual-stream hierarchical fusion engine network includes the following steps: A dual-stream hierarchical fusion encoder is constructed to collaboratively extract global and local features from SAR images at different time phases. The enhanced features are then used in a hierarchically aware U-shaped decoder to achieve hierarchical modeling and refined representation of the difference features. A frequency-domain channel attention mechanism with fused spatial weights is designed to enhance channel selectivity in the frequency domain and strengthen the response to change regions in the spatial dimension. A noise-resistant weighted loss function based on the difference map ablation coefficient is constructed to adaptively adjust the loss weights of each region, achieving noise region suppression and change region enhancement, effectively mitigating gradient bias caused by class imbalance and speckle noise. Training includes a dual-stream hierarchical fusion encoder and a U-shaped decoder network; SAR images from different time phases are input into the trained dual-stream hierarchical fusion engine network, which outputs a change detection binary map.

The invention provides a civil aircraft operation risk cause portraying method based on an adaptive multi-task neural network, and the method comprises the steps: firstly, building a civil aircraft operation risk cause portraying hierarchical label system, and providing a basic support for the comprehensive representation of primary risk causes covering people, aircrafts, rings and management and fine-grained secondary causes thereof; then, an instance hardness threshold sampling algorithm is introduced to relieve the problem of sample category imbalance caused by multi-category risks; and finally, constructing a civil aircraft operation risk cause portraying method based on an adaptive multi-task neural network, and designing an adaptive weighting strategy oriented to multi-task optimization to realize comprehensive and detailed risk cause description. According to the method, dynamic adaptive weighting of multiple tasks in the civil aircraft operation risk cause portrait model is realized, and intelligent auxiliary decision support can be provided for on-site potential safety hazard investigation and safety protection measure formulation.

The application discloses a kind of training sample selection methods for hyperspectral remote sensing image classification, belong to image classification technical field.The following steps are included: given unlabelled dataset, utilize pre-training GSCVIT model to extract classification features and multi-head attention weight, calculate spatial attention entropy and be spliced into enhanced features with classification features, through K-Center greedy algorithm screening core set sample subset, with the aid of group sampling loader loading data, using dynamic distribution balance loss (DDB Loss) training model to optimize classification performance.The application is verified on four datasets, and the results show that the method can effectively select key samples, dynamically optimize class distribution, significantly improve the classification accuracy and stability of the model under unbalanced data, and enhance the recognition ability of minority classes and weak class targets.Solve the problems of high labeling cost, class imbalance and complex feature expression of hyperspectral remote sensing image.

The invention discloses a rectification current bearing fault data generation method and system based on space-time condition fusion and non-classifier guidance, and belongs to the field of industrial fault diagnosis. The method comprises the following steps: converting a one-dimensional vibration signal into a two-dimensional time-frequency diagram through continuous wavelet transform, inputting the two-dimensional time-frequency diagram into a UNet network, and embedding a residual Seaformer module to fuse time-space condition information; a rectification flow path consistency loss function is adopted to train a network, a high-fidelity synthesis sample is generated through non-classifier guidance and Euler integration in the reasoning stage, and enhanced data is obtained through cosine similarity screening. According to the system, under the 1: 400 extreme unbalance condition, the accuracy rate is larger than or equal to 99% in a CWRU data set, the SEU data set is reduced to 81.15% due to complex working conditions, the reasoning time is only 6.7% of that of DDPM, and the system can be deployed on edge equipment and is suitable for scenes such as electric power, mines and rail transit.

This invention relates to a deep learning-based method for evaluating the fracturing effect of tight reservoirs, belonging to the technical field of tight reservoir fracturing effect evaluation. This deep learning-based method acquires full-wavelength data before and after fracturing using dipole acoustic logging. Preprocessing techniques such as bandpass filtering and normalization are used to "align" noise and scale differences caused by different well sections / instruments. WGAN-GP is used to generate adversarial examples for data augmentation to alleviate insufficient samples and class imbalance. Based on this, a residual network incorporating SE attention and models such as DenseNet are constructed, and mechanisms such as FocalLoss, weighted loss, regularization, and cosine annealing are employed to improve training stability and anti-overfitting ability, thereby achieving high-precision identification of fracture height and adaptation across well sections and instruments.

A computer based system and method for synthetic text generation includes a processor. The processor implements a text style transfer algorithm to first input data to generate complaints data from non-complaint emails data associated with a plurality non-complaint emails. The processor converts the plurality of non-complaint emails into a first set of complaint emails based on implementing the text style transfer algorithm and implements a text generation model algorithm to second input data to generate a second set of complaint emails from a plurality of complaint emails. The processor also generates a set of synthetic complaint emails based on the generated first set of complaint emails and the second set of complaint emails; trains a model based on the generated synthetic complaint emails; and applies the trained model to a new set of emails to resolve class imbalance in automatic complaint identification from the new set of emails.