Stacked What-Where Auto-encoders

TL;DR

The paper introduces SWWAE, a novel neural network architecture that combines discriminative and generative pathways for versatile learning without sampling, using a convolutional and deconvolutional network with a unique 'what-where' pooling mechanism.

Contribution

It proposes SWWAE, a unified model integrating supervised, semi-supervised, and unsupervised learning with a novel 'what-where' pooling approach and reconstruction-based training.

Findings

Effective in semi-supervised learning tasks

Achieves competitive results without sampling during training

Unifies discriminative and generative modeling in a single architecture

Abstract

We present a novel architecture, the "stacked what-where auto-encoders" (SWWAE), which integrates discriminative and generative pathways and provides a unified approach to supervised, semi-supervised and unsupervised learning without relying on sampling during training. An instantiation of SWWAE uses a convolutional net (Convnet) (LeCun et al. (1998)) to encode the input, and employs a deconvolutional net (Deconvnet) (Zeiler et al. (2010)) to produce the reconstruction. The objective function includes reconstruction terms that induce the hidden states in the Deconvnet to be similar to those of the Convnet. Each pooling layer produces two sets of variables: the "what" which are fed to the next layer, and its complementary variable "where" that are fed to the corresponding layer in the generative decoder.