Skip to yearly menu bar Skip to main content


Poster

On the loss landscape of a class of deep neural networks with no bad local valleys

Quynh Nguyen · Mahesh Chandra Mukkamala · Matthias Hein

Great Hall BC #24

Keywords: [ deep neural networks ] [ loss landscape ] [ local minima ]


Abstract:

We identify a class of over-parameterized deep neural networks with standard activation functions and cross-entropy loss which provably have no bad local valley, in the sense that from any point in parameter space there exists a continuous path on which the cross-entropy loss is non-increasing and gets arbitrarily close to zero. This implies that these networks have no sub-optimal strict local minima.

Live content is unavailable. Log in and register to view live content