Towards a General Theory of Infinite-Width Limits of Neural Classifiers
arXiv:2003.05884
Abstract
Obtaining theoretical guarantees for neural networks training appears to be a hard problem in a general case. Recent research has been focused on studying this problem in the limit of infinite width and two different theories have been developed: a mean-field (MF) and a constant kernel (NTK) limit theories. We propose a general framework that provides a link between these seemingly distinct theories. Our framework out of the box gives rise to a discrete-time MF limit which was not previously explored in the literature. We prove a convergence theorem for it and show that it provides a more reasonable approximation for finite-width nets compared to the NTK limit if learning rates are not very small. Also, our framework suggests a limit model that coincides neither with the MF limit nor with the NTK one. We show that for networks with more than two hidden layers RMSProp training has a non-trivial discrete-time MF limit but GD training does not have one. Overall, our framework demonstrates that both MF and NTK limits have considerable limitations in approximating finite-sized neural nets, indicating the need for designing more accurate infinite-width approximations for them.
arXiv comment27 pages, 7 figures, accepted to ICML'2020
References in corpus (9)
- Neural Tangent Kernel: Convergence and Generalization in Neural Networks
- A Mean Field View of the Landscape of Two-Layers Neural Networks
- Exponential expressivity in deep neural networks through transient chaos
- On Lazy Training in Differentiable Programming
- Mean-field theory of two-layers neural networks: dimension-free bounds and kernel limit
- A Mean Field Theory of Batch Normalization
- Mean Field Limit of the Learning Dynamics of Multilayer Neural Networks
- Beyond Linearization: On Quadratic and Higher-Order Approximation of Wide Neural Networks
- Collective evolution of weights in wide neural networks