ViTSGMM: A Robust Semi-Supervised Image Recognition Network Using Sparse Labels

2025-06-04

AI Summary

Image recognition under extremely sparse labeling—e.g., only 1–5 labeled samples per class—suffers from poor generalization. Method: We propose a hierarchical hybrid density classification framework grounded in mutual information maximization. It integrates a hierarchical sparse-label classification architecture, vision Transformer-based feature extraction, sparse Gaussian mixture modeling (SGMM), and feature redundancy compression. Contribution/Results: We identify and rectify a long-standing semi-supervised data leakage issue in the STL-10 benchmark. Extensive experiments demonstrate that our method achieves significant improvements over state-of-the-art approaches on STL-10, CIFAR-10, and CIFAR-100 under minimal labeling budgets (e.g., 1–5 samples per class). The implementation is publicly available.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text

Abstract

We present ViTSGMM, an image recognition network that leverages semi-supervised learning in a highly efficient manner. Existing works often rely on complex training techniques and architectures, while their generalization ability when dealing with extremely limited labeled data remains to be improved. To address these limitations, we construct a hierarchical mixture density classification decision mechanism by optimizing mutual information between feature representations and target classes, compressing redundant information while retaining crucial discriminative components. Experimental results demonstrate that our method achieves state-of-the-art performance on STL-10 and CIFAR-10/100 datasets when using negligible labeled samples. Notably, this paper also reveals a long-overlooked data leakage issue in the STL-10 dataset for semi-supervised learning tasks and removes duplicates to ensure the reliability of experimental results. Code available at https://github.com/Shu1L0n9/ViTSGMM.

Problem

• Improves semi-supervised image recognition with sparse labels
• Addresses limited labeled data generalization in existing methods
• Identifies and fixes data leakage in STL-10 dataset

Innovation

• Hierarchical mixture density classification decision mechanism
• Optimizing mutual information for feature representation
• Removing data leakage for reliable results