Neural Mechanics: Symmetry and Broken Conservation Laws in Deep Learning Dynamics
Preprint
Abstract
Understanding the dynamics of neural network parameters during training is one of the key challenges in building a theoretical foundation for deep learning. A central obstacle is that the motion of a network in high-dimensional parameter space undergoes discrete finite steps along complex stochastic gradients derived from real-world datasets. We circumvent this obstacle through a unifying theoretical framework based on intrinsic symmetries embedded in a network's architecture that are present for any dataset. We show that any such symmetry imposes stringent geometric constraints on gradients and Hessians, leading to an associated conservation law in the continuous-time limit of stochastic gradient descent (SGD), akin to Noether's theorem in physics. We further show that finite learning rates used in practice can actually break these symmetry induced conservation laws. We apply tools from finite difference methods to derive modified gradient flow, a differential equation that better approximates the numerical trajectory taken by SGD at finite learning rates. We combine modified gradient flow with our framework of symmetries to derive exact integral expressions for the dynamics of certain parameter combinations. We empirically validate our analytic expressions for learning dynamics on VGG-16 trained on Tiny ImageNet. Overall, by exploiting symmetry, our work demonstrates that we can analytically describe the learning dynamics of various parameter combinations at finite learning rates and batch sizes for state of the art architectures trained on any dataset.Keywords
All Related Versions
This publication has 6 references indexed in Scilit:
- Uniform-in-time weak error analysis for stochastic gradient descent algorithms via diffusion approximationCommunications in Mathematical Sciences, 2020
- A mean field view of the landscape of two-layer neural networksProceedings of the National Academy of Sciences, 2018
- On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance PropagationPLOS ONE, 2015
- Stochastic Approximation and Recursive Algorithms and Applications2003
- Neural networks and principal component analysis: Learning from examples without local minimaNeural Networks, 1989
- The modified equation approach to the stability and accuracy analysis of finite-difference methodsJournal of Computational Physics, 1974