Long-term Reproducibility for Neural Architecture Search

TL;DR

This paper highlights the importance of long-term reproducibility in Neural Architecture Search (NAS), proposing a checklist to improve reproducibility and evaluating its effectiveness across various NAS methods.

Contribution

It introduces a comprehensive checklist for ensuring long-term reproducibility in NAS research, addressing gaps in existing reproducibility efforts.

Findings

Checklist improves reproducibility standards

Retrospective application enhances existing NAS approaches

Highlights long-term reproducibility challenges in NAS

Abstract

It is a sad reflection of modern academia that code is often ignored after publication -- there is no academic 'kudos' for bug fixes / maintenance. Code is often unavailable or, if available, contains bugs, is incomplete, or relies on out-of-date / unavailable libraries. This has a significant impact on reproducibility and general scientific progress. Neural Architecture Search (NAS) is no exception to this, with some prior work in reproducibility. However, we argue that these do not consider long-term reproducibility issues. We therefore propose a checklist for long-term NAS reproducibility. We evaluate our checklist against common NAS approaches along with proposing how we can retrospectively make these approaches more long-term reproducible.