0

Contrastive Embeddings for Neural Architectures

Contrastive learning generates architecture embeddings, enabling black-box optimization, state-of-the-art performance in Neural Architecture Search, and transfer learning between search spaces.

Year
2021
Venue
contrastive-embeddings-for-neural-1
Authors
2
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2102.04208v2ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

The performance of algorithms for neural architecture search strongly depends on the parametrization of the search space. We use contrastive learning to identify networks across different initializations based on their data Jacobians, and automatically produce the first architecture embeddings independent from the parametrization of the search space. Using our contrastive embeddings, we show that traditional black-box optimization algorithms, without modification, can reach state-of-the-art performance in Neural Architecture Search. As our method provides a unified embedding space, we perform for the first time transfer learning between search spaces. Finally, we show the evolution of embeddings during training, motivating future studies into using embeddings at different training stages to gain a deeper understanding of the networks in a search space.

Authors

2