A robust width learning image classification model construction and training method for denoising and structure enhancement

CN122265735APending Publication Date: 2026-06-23NANJING MEDICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NANJING MEDICAL UNIV
Filing Date
2026-04-29
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing width learning systems have limitations in label modeling for image classification, making it difficult to fully reflect the true relationships between samples, and redundant features and noise affect model performance.

Method used

The classification model is optimized by using the alternating direction multiplier method. By combining ridge regression, graph regularization and matrix factorization, the discriminative power of feature representation is improved and noise interference is suppressed through various structured constraints.

Benefits of technology

It effectively improves the discriminative power and stability of the model's feature representation, reduces redundant information, enhances intra-class consistency and inter-class discrimination, and improves the accuracy and generalization ability of image classification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122265735A_ABST
    Figure CN122265735A_ABST
Patent Text Reader

Abstract

The application discloses a denoising and structured enhancement robust width learning image classification model construction and training method, comprising the following steps: S1, setting a training set and a test set, and preprocessing input data; S2, using a width learning system to map and enhance the training set data after preprocessing in S1, and expanding the training set data into high-dimensional generalized feature representation; S3, on the basis of the high-dimensional feature representation in S2, a classification model based on the width learning system is constructed; S4, using an alternating direction multiplier method to block optimize the classification model in S3, and iteratively solving an optimal projection matrix to ensure the convergence and optimization stability of the model under the constraint condition; and S5, on the basis of the optimal projection matrix in S4, the classification model in S3 is used to classify and predict the test set data after preprocessing in S1.
Need to check novelty before this filing date? Find Prior Art