The Lingering of Gradients: Theory and Applications

TL;DR

This paper introduces a refined analysis of gradient-based methods by considering the lingering effect of gradients, leading to faster convergence rates and improved practical performance in large-scale optimization tasks.

Contribution

It develops a theoretical framework for gradient lingering, demonstrating improved convergence rates and applying it to real-world large-scale problems.

Findings

Gradient descent convergence rate improved from 1/T to exp(-T^{1/3})

Achieved high-accuracy solutions on large-scale datasets with fewer passes

Enhanced SVM performance by two orders of magnitude over existing algorithms

Abstract

Classically, the time complexity of a first-order method is estimated by its number of gradient computations. In this paper, we study a more refined complexity by taking into account the `lingering' of gradients: once a gradient is computed at $x_k$, the additional time to compute gradients at $x_{k+1},x_{k+2},\dots$ may be reduced. We show how this improves the running time of several first-order methods. For instance, if the `additional time' scales linearly with respect to the traveled distance, then the `convergence rate' of gradient descent can be improved from $1/T$ to $\exp(-T^{1/3})$. On the application side, we solve a hypothetical revenue management problem on the Yahoo! Front Page Today Module with 4.6m users to $10^{-6}$ error using only 6 passes of the dataset; and solve a real-life support vector machine problem to an accuracy that is two orders of magnitude better comparing to the state-of-the-art algorithm.