Location : Seminar Room 638, Institute of Mathematics (NTU Campus)
Speaker : Govind Menon (Brown University)
Organizer : Colin McSwiggen (AS)
Abstract : This lecture describes the surprising mathematical structure of the Deep Linear Network (DLN), a restriction of deep learning to linear functions. We then use the DLN to formulate a geometric picture of function approximation by (nonlinear) feed-forward neural networks.