Faster Adaptive Optimization via Expected Gradient Outer Product Reparameterization
This addresses a foundational issue in machine learning optimization by potentially enhancing the performance of widely used algorithms like Adam and Adagrad, though it appears incremental as it builds on existing adaptive methods.
The paper tackles the problem of adaptive optimization algorithms' sensitivity to parameterization by proposing a reparameterization method based on the expected gradient outer product (EGOP) matrix, showing theoretically and empirically that it can improve convergence behavior for a broad class of functions.
Adaptive optimization algorithms -- such as Adagrad, Adam, and their variants -- have found widespread use in machine learning, signal processing and many other settings. Several methods in this family are not rotationally equivariant, meaning that simple reparameterizations (i.e. change of basis) can drastically affect their convergence. However, their sensitivity to the choice of parameterization has not been systematically studied; it is not clear how to identify a "favorable" change of basis in which these methods perform best. In this paper we propose a reparameterization method and demonstrate both theoretically and empirically its potential to improve their convergence behavior. Our method is an orthonormal transformation based on the expected gradient outer product (EGOP) matrix, which can be approximated using either full-batch or stochastic gradient oracles. We show that for a broad class of functions, the sensitivity of adaptive algorithms to choice-of-basis is influenced by the decay of the EGOP matrix spectrum. We illustrate the potential impact of EGOP reparameterization by presenting empirical evidence and theoretical arguments that common machine learning tasks with "natural" data exhibit EGOP spectral decay.