Scalable Transfer Learning with Expert Models

TL;DR

This paper presents a scalable transfer learning method using expert models, which efficiently selects relevant experts for new tasks, significantly reducing computation and outperforming baselines on diverse vision tasks.

Contribution

The paper introduces a novel expert-based transfer learning strategy that scales efficiently without revisiting pre-training data and compresses multiple experts into a single adaptable model.

Findings

Achieves 2-3 orders of magnitude speed-up over existing methods.

Outperforms baseline models on over 20 diverse vision tasks.

Demonstrates effective expert selection using performance proxies.

Abstract

Transfer of pre-trained representations can improve sample efficiency and reduce computational requirements for new tasks. However, representations used for transfer are usually generic, and are not tailored to a particular distribution of downstream tasks. We explore the use of expert representations for transfer with a simple, yet effective, strategy. We train a diverse set of experts by exploiting existing label structures, and use cheap-to-compute performance proxies to select the relevant expert for each target task. This strategy scales the process of transferring to new tasks, since it does not revisit the pre-training data during transfer. Accordingly, it requires little extra compute per target task, and results in a speed-up of 2-3 orders of magnitude compared to competing approaches. Further, we provide an adapter-based architecture able to compress many experts into a single model. We evaluate our approach on two different data sources and demonstrate that it outperforms baselines on over 20 diverse vision tasks in both cases.