GOLD-NAS: Gradual, One-Level, Differentiable

TL;DR

GOLD-NAS introduces a novel differentiable neural architecture search method that enlarges the search space and gradually prunes weak operators, efficiently discovering Pareto-optimal architectures with improved tradeoffs between accuracy and complexity.

Contribution

The paper proposes GOLD-NAS, a new algorithm that relaxes constraints and employs a resource-aware, gradual pruning approach for more effective neural architecture search.

Findings

Enlarged search space with over 10^160 candidates.

GOLD-NAS finds Pareto-optimal architectures efficiently.

Discovered architectures outperform previous methods in accuracy-complexity tradeoff.

Abstract

There has been a large literature of neural architecture search, but most existing work made use of heuristic rules that largely constrained the search flexibility. In this paper, we first relax these manually designed constraints and enlarge the search space to contain more than $10^{160}$ candidates. In the new space, most existing differentiable search methods can fail dramatically. We then propose a novel algorithm named Gradual One-Level Differentiable Neural Architecture Search (GOLD-NAS) which introduces a variable resource constraint to one-level optimization so that the weak operators are gradually pruned out from the super-network. In standard image classification benchmarks, GOLD-NAS can find a series of Pareto-optimal architectures within a single search procedure. Most of the discovered architectures were never studied before, yet they achieve a nice tradeoff between recognition accuracy and model complexity. We believe the new space and search algorithm can advance the search of differentiable NAS.