Abstract:Image classification by semi-supervised learning has recently become a hot spot, and the Co-Training framework is an important method of semi-supervised image classification. In the traditional Co-Training structure, the sub-networks will generate pseudo-labels for each other, and these pseudo-labels will further be used as a supervisory signal for model training. However, the pseudo-labels will hurt classification performance because of their low accuracy and unbalanced distribution. In this article, we are trying to solve the preceding two problems by designing the Balanced Module (BM) and Gaussian Mixture Module (GMM), and propose BAPS (the B alanced and A ccurate P seudo-labels for S emi-supervised image classification). In BM, the two sub-networks jointly predict the unlabeled images, then select the pseudo-labels with a high-confidence threshold to perform the balancing operation to obtain the initial samples with balanced distribution of each category. In GMM, referring to the common practice of the Learning from Noise Labels task, we use GMM to fit the loss distribution of images with pseudo-labels output by BM, then clean samples and noise samples are divided based on the observation that the loss of correctly labeled images is generally smaller than that of wrongly labeled ones. Through BM and GMM, pseudo-labels with balanced distribution and high accuracy are obtained for the subsequent model training process. Our model has achieved better classification accuracy than most state-of-the-art semi-supervised image classification algorithms on the CIFAR-10/100 and SVHN datasets, and further ablation experiments demonstrate the effectiveness of our BAPS. The source code of BAPS will be available at https://github.com/zhaojianaaa .

A Hybrid Generative/discriminative Method for Semi-Supervised Classification

Dual-Classifier Collaborative Method Based on Semi-Supervised Active Learning

Hybrid Generative-Discriminative Classification Using Posterior Divergence

Development of Co-Training Support Vector Machine Model for Semi-Supervised Classification

Optimally Combining Classifiers for Semi-Supervised Learning

Semi-supervised Learning Combining Co-Training with Active Learning

A Hybrid of Generative and Discriminative Models Based on the Gaussian-coupled Softmax Layer

A Hybrid Discriminative Approach with Bayesian Prior Constraint

One of semi-supervised classification algorithm named Co-S3OM based on cooperative training

Combining deep generative and discriminative models for Bayesian semi-supervised learning

Multi-Train: A Semi-Supervised Heterogeneous Ensemble Classifier

Semi-Supervised Learning with Flexible Discriminator Objective in Generative Adversarial Networks Framework

Semi-supervised regression via embedding space mapping and pseudo-label smearing

Exploiting Ensemble Method in Semi-Supervised Learning

Conditional Generative Adversarial Network on Semi-supervised Learning Task

A Generative/Discriminative Hybrid Model: Bayes Perceptron Classifier

Semi-Supervised Learning with Multi-Head Co-Training

A hybrid approach to sentiment classification and feature expansion strategy

Balanced and Accurate Pseudo-Labels for Semi-Supervised Image Classification

Semi-Supervised Hybrid Clustering by Integrating Gaussian Mixture Model and Distance Metric Learning

Semi-Supervised Ensemble Classification Method Based On Near Neighbor And Its Application