Lookaround Optimizer: $k$ steps around, 1 step average

TL;DR

Lookaround is a novel SGD-based optimizer that iteratively trains multiple networks with data augmentation and averages them, leading to flatter minima and improved generalization in deep learning models.

Contribution

It introduces a new optimizer combining around and average steps, enhancing diversity and weight locality for better training outcomes.

Findings

Outperforms state-of-the-art methods on CIFAR and ImageNet benchmarks.

Effective for both CNNs and ViTs.

Theoretically justified by convergence analysis.

Abstract

Weight Average (WA) is an active research topic due to its simplicity in ensembling deep networks and the effectiveness in promoting generalization. Existing weight average approaches, however, are often carried out along only one training trajectory in a post-hoc manner (i.e., the weights are averaged after the entire training process is finished), which significantly degrades the diversity between networks and thus impairs the effectiveness. In this paper, inspired by weight average, we propose Lookaround, a straightforward yet effective SGD-based optimizer leading to flatter minima with better generalization. Specifically, Lookaround iterates two steps during the whole training period: the around step and the average step. In each iteration, 1) the around step starts from a common point and trains multiple networks simultaneously, each on transformed data by a different data augmentation, and 2) the average step averages these trained networks to get the averaged network, which serves as the starting point for the next iteration. The around step improves the functionality diversity while the average step guarantees the weight locality of these networks during the whole training, which is essential for WA to work. We theoretically explain the superiority of Lookaround by convergence analysis, and make extensive experiments to evaluate Lookaround on popular benchmarks including CIFAR and ImageNet with both CNNs and ViTs, demonstrating clear superiority over state-of-the-arts. Our code is available at https://github.com/Ardcy/Lookaround.