Deep Learning by Scattering

Abstract

We introduce general scattering transforms as mathematical models of deep neural networks with l2 pooling. Scattering networks iteratively apply complex valued unitary operators, and the pooling is performed by a complex modulus. An expected scattering defines a contractive representation of a high-dimensional probability distribution, which preserves its mean-square norm. We show that unsupervised learning can be casted as an optimization of the space contraction to preserve the volume occupied by unlabeled examples, at each layer of the network. Supervised learning and classification are performed with an averaged scattering, which provides scattering estimations for multiple classes.

10 pages, 1 figure

Cited by in corpus (6)

• A Deep Representation for Invariance And Music Classification
• Building a Regular Decision Boundary with Deep Networks
• Learning to compress and search visual data in large-scale systems
• Learning Non-Parametric Invariances from Data with Permanent Random Connectomes
• Deep Haar Scattering Networks
• Interferometric Graph Transform for Community Labeling