|Location:||MSRI: Baker Board Room|
Title: A harmonic analysis perspective on deep learning
Abstract: Deep convolutional neural networks have become one of the most powerful types of machine learning tools, yet there is no adequate theory that explains their remarkable performance. Motivated by this problem, Mallat recently discovered a connection between harmonic analysis and neural networks. In this talk, we illustrate the motivation for his wavelet-based theory and outline his main results. Further, we shall discuss an alternative Fourier-based approach to deep learning, potential applications, and directions for further research. This is joint work with Wojciech Czaja.