Stretchable Cells Help DARTS Search Better

TL;DR

This paper introduces stretchable cells in DARTS to improve search quality by explicitly modeling topology, leading to better performance with fewer layers and parameters on CIFAR-10 and ImageNet.

Contribution

It proposes a novel stretchable cell design with topological variables, enabling simultaneous operation and topology search, addressing topology collapse issues in DARTS.

Findings

Improved DARTS accuracy by 0.28% on CIFAR-10 with fewer parameters.

Achieved 2.9% accuracy gain on ImageNet with similar FLOPs.

Enhanced adaptability to various layer numbers and topologies.

Abstract

Differentiable neural architecture search (DARTS) has gained much success in discovering flexible and diverse cell types. To reduce the evaluation gap, the supernet is expected to have identical layers with the target network. However, even for this consistent search, the searched cells often suffer from poor performance, especially for the supernet with fewer layers, as current DARTS methods are prone to wide and shallow cells, and this topology collapse induces sub-optimal searched cells. In this paper, we alleviate this issue by endowing the cells with explicit stretchability, so the search can be directly implemented on our stretchable cells for both operation type and topology simultaneously. Concretely, we introduce a set of topological variables and a combinatorial probabilistic distribution to explicitly model the target topology. With more diverse and complex topologies, our method adapts well for various layer numbers. Extensive experiments on CIFAR-10 and ImageNet show that our stretchable cells obtain better performance with fewer layers and parameters. For example, our method can improve DARTS by 0.28\% accuracy on CIFAR-10 dataset with 45\% parameters reduced or 2.9\% with similar FLOPs on ImageNet dataset.