Learning by Association - A versatile semi-supervised training method for neural networks

TL;DR

This paper introduces a semi-supervised training method for neural networks that leverages associations between labeled and unlabeled data to improve classification performance, especially with limited labeled data.

Contribution

The paper presents a novel association-based semi-supervised training framework inspired by human learning, easily integrable into existing neural network training pipelines.

Findings

Outperforms state-of-the-art on SVHN with few labeled samples

Significantly improves classification accuracy using unlabeled data

Easy to implement and adaptable to various neural network architectures

Abstract

In many real-world scenarios, labeled data for a specific machine learning task is costly to obtain. Semi-supervised training methods make use of abundantly available unlabeled data and a smaller number of labeled examples. We propose a new framework for semi-supervised training of deep neural networks inspired by learning in humans. "Associations" are made from embeddings of labeled samples to those of unlabeled ones and back. The optimization schedule encourages correct association cycles that end up at the same class from which the association was started and penalizes wrong associations ending at a different class. The implementation is easy to use and can be added to any existing end-to-end training setup. We demonstrate the capabilities of learning by association on several data sets and show that it can improve performance on classification tasks tremendously by making use of additionally available unlabeled data. In particular, for cases with few labeled data, our training scheme outperforms the current state of the art on SVHN.