AET vs. AED: Unsupervised Representation Learning by Auto-Encoding Transformations rather than Data

TL;DR

This paper introduces a new unsupervised learning paradigm called Auto-Encoding Transformation (AET), which predicts transformations from encoded features, leading to improved representations that outperform existing methods on standard datasets.

Contribution

The paper proposes AET, a novel unsupervised learning approach that encodes visual structures to predict transformations, expanding the scope of transformations beyond traditional data auto-encoding.

Findings

AET achieves state-of-the-art results on CIFAR-10, ImageNet, and Places datasets.

AET outperforms existing unsupervised methods significantly.

AET approaches the performance of fully supervised models.

Abstract

The success of deep neural networks often relies on a large amount of labeled examples, which can be difficult to obtain in many real scenarios. To address this challenge, unsupervised methods are strongly preferred for training neural networks without using any labeled data. In this paper, we present a novel paradigm of unsupervised representation learning by Auto-Encoding Transformation (AET) in contrast to the conventional Auto-Encoding Data (AED) approach. Given a randomly sampled transformation, AET seeks to predict it merely from the encoded features as accurately as possible at the output end. The idea is the following: as long as the unsupervised features successfully encode the essential information about the visual structures of original and transformed images, the transformation can be well predicted. We will show that this AET paradigm allows us to instantiate a large variety of transformations, from parameterized, to non-parameterized and GAN-induced ones. Our experiments show that AET greatly improves over existing unsupervised approaches, setting new state-of-the-art performances being greatly closer to the upper bounds by their fully supervised counterparts on CIFAR-10, ImageNet and Places datasets.