Efficient NAS with FaDE on Hierarchical Spaces

TL;DR

FaDE introduces a differentiable approach for neural architecture search in hierarchical spaces, enabling efficient exploration and ranking of architectures without proxy spaces, demonstrated to correlate well with actual performance.

Contribution

FaDE presents a novel differentiable ranking method for hierarchical NAS that reduces exploration cost and improves performance prediction accuracy.

Findings

FaDE ranks correlate with actual architecture performance.

FaDE enables efficient exploration of deep hierarchical spaces.

Pseudo-gradient evolutionary search improves NAS efficiency.

Abstract

Neural architecture search (NAS) is a challenging problem. Hierarchical search spaces allow for cheap evaluations of neural network sub modules to serve as surrogate for architecture evaluations. Yet, sometimes the hierarchy is too restrictive or the surrogate fails to generalize. We present FaDE which uses differentiable architecture search to obtain relative performance predictions on finite regions of a hierarchical NAS space. The relative nature of these ranks calls for a memory-less, batch-wise outer search algorithm for which we use an evolutionary algorithm with pseudo-gradient descent. FaDE is especially suited on deep hierarchical, respectively multi-cell search spaces, which it can explore by linear instead of exponential cost and therefore eliminates the need for a proxy search space. Our experiments show that firstly, FaDE-ranks on finite regions of the search space correlate with corresponding architecture performances and secondly, the ranks can empower a pseudo-gradient evolutionary search on the complete neural architecture search space.