Stacked What-Where Auto-encoders

About

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.

Junbo Zhao, Michael Mathieu, Ross Goroshin, Yann LeCun · 2015

Related benchmarks

Task	Dataset	Result
Image Classification	MNIST (test)	Accuracy—	894
Image Classification	CIFAR-100	Accuracy69.12	691
Image Classification	CIFAR-10	Accuracy92.23	564
Image Classification	SVHN (test)	Accuracy—	470
Image Classification	STL-10 (test)	Accuracy74.33	380
Image Clustering	CIFAR-10	NMI0.233	318
Image Clustering	STL-10	ACC27	282
Classification	SVHN (test)	Error Rate23.56	182
Image Classification	STL-10	Accuracy74.3	129
Image Clustering	CIFAR-100	ACC14.7	111