Novelty Driven Evolutionary Neural Architecture Search

TL;DR

This paper introduces NEvoNAS, a multi-objective evolutionary approach that balances architecture novelty and accuracy, effectively avoiding local optima and reducing computational costs in neural architecture search.

Contribution

It proposes a novel multi-objective NAS method combining novelty search with supernet-based fitness estimation, improving search diversity and efficiency.

Findings

NEvoNAS outperforms previous EA-based NAS methods on two search spaces.

The method achieves better architectures with less computational resources.

Maintaining diversity via novelty search helps avoid local optima.

Abstract

Evolutionary algorithms (EA) based neural architecture search (NAS) involves evaluating each architecture by training it from scratch, which is extremely time-consuming. This can be reduced by using a supernet for estimating the fitness of an architecture due to weight sharing among all architectures in the search space. However, the estimated fitness is very noisy due to the co-adaptation of the operations in the supernet which results in NAS methods getting trapped in local optimum. In this paper, we propose a method called NEvoNAS wherein the NAS problem is posed as a multi-objective problem with 2 objectives: (i) maximize architecture novelty, (ii) maximize architecture fitness/accuracy. The novelty search is used for maintaining a diverse set of solutions at each generation which helps avoiding local optimum traps while the architecture fitness is calculated using supernet. NSGA-II is used for finding the \textit{pareto optimal front} for the NAS problem and the best architecture in the pareto front is returned as the searched architecture. Exerimentally, NEvoNAS gives better results on 2 different search spaces while using significantly less computational resources as compared to previous EA-based methods. The code for our paper can be found in https://github.com/nightstorm0909/NEvoNAS.