Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets

TL;DR

This paper introduces MetaD2A, a meta-learning framework that rapidly generates neural network architectures for new datasets by learning from a database of datasets and pretrained models, significantly reducing search time.

Contribution

The paper presents MetaD2A, a novel meta-learning approach that generalizes neural architecture search across multiple datasets with minimal computational cost.

Findings

MetaD2A generalizes well to unseen datasets like CIFAR-10 and CIFAR-100.

MetaD2A achieves a search time of approximately 33 GPU seconds.

MetaD2A outperforms existing methods like NSGANetV2 in speed with comparable accuracy.

Abstract

Despite the success of recent Neural Architecture Search (NAS) methods on various tasks which have shown to output networks that largely outperform human-designed networks, conventional NAS methods have mostly tackled the optimization of searching for the network architecture for a single task (dataset), which does not generalize well across multiple tasks (datasets). Moreover, since such task-specific methods search for a neural architecture from scratch for every given task, they incur a large computational cost, which is problematic when the time and monetary budget are limited. In this paper, we propose an efficient NAS framework that is trained once on a database consisting of datasets and pretrained networks and can rapidly search for a neural architecture for a novel dataset. The proposed MetaD2A (Meta Dataset-to-Architecture) model can stochastically generate graphs (architectures) from a given set (dataset) via a cross-modal latent space learned with amortized meta-learning. Moreover, we also propose a meta-performance predictor to estimate and select the best architecture without direct training on target datasets. The experimental results demonstrate that our model meta-learned on subsets of ImageNet-1K and architectures from NAS-Bench 201 search space successfully generalizes to multiple unseen datasets including CIFAR-10 and CIFAR-100, with an average search time of 33 GPU seconds. Even under MobileNetV3 search space, MetaD2A is 5.5K times faster than NSGANetV2, a transferable NAS method, with comparable performance. We believe that the MetaD2A proposes a new research direction for rapid NAS as well as ways to utilize the knowledge from rich databases of datasets and architectures accumulated over the past years. Code is available at https://github.com/HayeonLee/MetaD2A.