Local minima in training of deep networks

Grzegorz Swirszcz, Wojciech Marian Czarnecki, Razvan Pascanu

Nov 04, 2016 (modified: Jan 12, 2017) ICLR 2017 conference submission readers: everyone
  • Abstract: There has been a lot of recent interest in trying to characterize the error surface of deep models. This stems from a long standing question. Given that deep networks are highly nonlinear systems optimized by local gradient methods, why do they not seem to be affected by bad local minima? It is widely believed that training of deep models using gradient methods works so well because the error surface either has no local minima, or if they exist they need to be close in value to the global minimum. It is known that such results hold under strong assumptions which are not satisfied by real models. In this paper we present examples showing that for such theorem to be true additional assumptions on the data, initialization schemes and/or the model classes have to be made. We look at the particular case of finite size datasets. We demonstrate that in this scenario one can construct counter-examples (datasets or initialization schemes) when the network does become susceptible to bad local minima over the weight space.
  • TL;DR: As a contribution to the discussion about error surface and the question why "deep and cheap" learning works so well we present concrete examples of local minima and obstacles arising in the training of deep models.
  • Conflicts: google.com
  • Keywords: Theory, Deep learning, Supervised Learning, Optimization