The Limiting Dynamics of SGD: Modified Loss, Phase Space Oscillations, and Anomalous Diffusion

Abstract

In this work we explore the limiting dynamics of deep neural networks trained with stochastic gradient descent (SGD). As observed previously, long after performance has converged, networks continue to move through parameter space by a process of anomalous diffusion in which distance travelled grows as a power law in the number of gradient updates with a nontrivial exponent. We reveal an intricate interaction between the hyperparameters of optimization, the structure in the gradient noise, and the Hessian matrix at the end of training that explains this anomalous diffusion. To build this understanding, we first derive a continuous-time model for SGD with finite learning rates and batch sizes as an underdamped Langevin equation. We study this equation in the setting of linear regression, where we can derive exact, analytic expressions for the phase space dynamics of the parameters and their instantaneous velocities from initialization to stationarity. Using the Fokker-Planck equation, we show that the key ingredient driving these dynamics is not the original training loss, but rather the combination of a modified loss, which implicitly regularizes the velocity, and probability currents, which cause oscillations in phase space. We identify qualitative and quantitative predictions of this theory in the dynamics of a ResNet-18 model trained on ImageNet. Through the lens of statistical physics, we uncover a mechanistic origin for the anomalous limiting dynamics of deep neural networks trained with SGD.

78 pages, 9 figures, Neural Computation 2024

References in corpus (11)

• One weird trick for parallelizing convolutional neural networks
• Gradient Descent Happens in a Tiny Subspace
• An Investigation into Neural Net Optimization via Hessian Eigenvalue Density
• On the Origin of Implicit Regularization in Stochastic Gradient Descent
• Stochasticity helps to navigate rough landscapes: comparing gradient-descent-based algorithms in the phase retrieval problem
• A Tail-Index Analysis of Stochastic Gradient Noise in Deep Neural Networks
• Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability
• On the Validity of Modeling SGD with Stochastic Differential Equations (SDEs)
• Neural Mechanics: Symmetry and Broken Conservation Laws in Deep Learning Dynamics
• Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate
• Anomalous diffusion dynamics of learning in deep neural networks

Cited by in corpus (3)

• Weight fluctuations in (deep) linear neural networks and a derivation of the inverse-variance flatness relation
• Noise Balance and Stationary Distribution of Stochastic Gradient Descent
• Anti-Correlated Noise in Epoch-Based Stochastic Gradient Descent: Implications for Weight Variances in Flat Directions