Dynamic Ensemble of Low-fidelity Experts: Mitigating NAS "Cold-Start"

TL;DR

This paper introduces a dynamic ensemble predictor framework that leverages multiple low-fidelity estimations to mitigate the cold-start problem in predictor-based Neural Architecture Search, improving prediction accuracy with less data.

Contribution

It proposes a novel two-step method to fuse diverse low-fidelity information using specialized experts and a gating network, enhancing NAS efficiency.

Findings

Effective across five search spaces with different encoders

Improves prediction accuracy with limited high-fidelity data

Easily integrated into existing NAS frameworks

Abstract

Predictor-based Neural Architecture Search (NAS) employs an architecture performance predictor to improve the sample efficiency. However, predictor-based NAS suffers from the severe ``cold-start'' problem, since a large amount of architecture-performance data is required to get a working predictor. In this paper, we focus on exploiting information in cheaper-to-obtain performance estimations (i.e., low-fidelity information) to mitigate the large data requirements of predictor training. Despite the intuitiveness of this idea, we observe that using inappropriate low-fidelity information even damages the prediction ability and different search spaces have different preferences for low-fidelity information types. To solve the problem and better fuse beneficial information provided by different types of low-fidelity information, we propose a novel dynamic ensemble predictor framework that comprises two steps. In the first step, we train different sub-predictors on different types of available low-fidelity information to extract beneficial knowledge as low-fidelity experts. In the second step, we learn a gating network to dynamically output a set of weighting coefficients conditioned on each input neural architecture, which will be used to combine the predictions of different low-fidelity experts in a weighted sum. The overall predictor is optimized on a small set of actual architecture-performance data to fuse the knowledge from different low-fidelity experts to make the final prediction. We conduct extensive experiments across five search spaces with different architecture encoders under various experimental settings. Our method can easily be incorporated into existing predictor-based NAS frameworks to discover better architectures.