Abstract
Deep learning has achieved unprecedented practical success in many applications. Despite its empirical success, however, the theoretical understanding of deep neural networks still remains a major open problem. In this paper, we explore properties of two-layered ReLU networks. For simplicity, we assume that the optimal model parameters (also called ground-truth parameters) are known. We then assume that a network receives Gaussian input and is trained by minimizing the expected squared loss between the prediction function of the network and a target function. To conduct the analysis, we propose a normal equation for critical points, and study the invariances under three kinds of transformations, namely, scale transformation, rotation transformation and perturbation transformation. We prove that these transformations can keep the loss of a critical point invariant, thus can incur flat regions. Consequently, how to escape from flat regions is vital in training neural networks.
Original language | English |
---|---|
Pages (from-to) | 545-560 |
Number of pages | 16 |
Journal | Journal of Machine Learning Research |
Volume | 77 |
Publication status | Published - 2017 |
Externally published | Yes |
Event | 9th Asian Conference on Machine Learning, ACML 2017 - Seoul, Korea, Republic of Duration: 15 Nov 2017 → 17 Nov 2017 |
Keywords
- Critical points
- Flatness
- Loss surface
- Two-layered ReLU network