Comparing Dynamics: Deep Neural Networks versus Glassy Systems

We analyze numerically the training dynamics of deep neural networks (DNN) by using methods developed in statistical physics of glassy systems. The two main issues we address are the complexity of the loss-landscape and of the dynamics within it, and to what extent DNNs share similarities with glassy systems. Our findings, obtained for different architectures and data-sets, suggest that during the training process the dynamics slows down because of an increasingly large number of flat directions. At large times, when the loss is approaching zero, the system diffuses at the bottom of the landscape. Despite some similarities with the dynamics of mean-field glassy systems, in particular, the absence of barrier crossing, we find distinctive dynamical behaviors in the two cases, thus showing that the statistical properties of the corresponding loss and energy landscapes are different. In contrast, when the network is under-parametrized we observe a typical glassy behavior, thus suggesting the existence of different phases depending on whether the network is under-parametrized or over-parametrized.


Published in:
Proceedings of the 35th International Conference on Machine Learning, PMLR 80, 314-323
Presented at:
35th International Conference on Machine Learning (ICML), Stockholm, Sweden, July 10-15, 2018
Year:
2018
Additional link:
Laboratories:




 Record created 2019-08-27, last modified 2019-08-30


Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)