Stacked Capsule Autoencoders

Adam Roman Kosiorek
Yee Whye Teh
Geoffrey Hinton
NeurIPS (2019)

Abstract

An object can be seen as a geometrically organized set of interrelated parts. A system that makes explicit use of these geometric relationships to recognize objects should be naturally robust to changes in viewpoint, because the intrinsic geometric relationships are viewpoint-invariant. We describe an unsupervised version of capsule networks, in which a neural encoder, which looks at all of the parts, is used to infer the presence and poses of object capsules. The encoder is trained by backpropagating through a decoder, which predicts the pose of each already discovered part using a mixture of pose predictions. The parts are discovered directly from an image, in a similar manner, by using a neural encoder, which infers parts and their affine transformations. The corresponding decoder models each image pixel as a mixture of predictions made by affine-transformed parts. We learn object- and their part-capsules on unlabeled data, and then cluster the vectors of presences of object capsules. When told the names of these clusters, we achieve state-of-the-art results for unsupervised classification on SVHN (55%) and near state-of-the-art on MNIST (98.5%).