Understanding the Transferability of Representations via Task-Relatedness

TL;DR

This paper introduces a task-relatedness metric to predict transfer learning success, demonstrating its effectiveness across vision and language tasks and enabling better model selection without requiring target task labels.

Contribution

It proposes a novel, computable measure of task-relatedness that explains transferability and aids in selecting pre-trained models for downstream tasks.

Findings

Task-relatedness correlates strongly with transfer performance.

The metric predicts transferability without needing target labels.

Using task-relatedness improves model selection for transfer learning.

Abstract

The growing popularity of transfer learning, due to the availability of models pre-trained on vast amounts of data, makes it imperative to understand when the knowledge of these pre-trained models can be transferred to obtain high-performing models on downstream target tasks. However, the exact conditions under which transfer learning succeeds in a cross-domain cross-task setting are still poorly understood. To bridge this gap, we propose a novel analysis that analyzes the transferability of the representations of pre-trained models to downstream tasks in terms of their relatedness to a given reference task. Our analysis leads to an upper bound on transferability in terms of task-relatedness, quantified using the difference between the class priors, label sets, and features of the two tasks. Our experiments using state-of-the-art pre-trained models show the effectiveness of task-relatedness in explaining transferability on various vision and language tasks. The efficient computability of task-relatedness even without labels of the target task and its high correlation with the model's accuracy after end-to-end fine-tuning on the target task makes it a useful metric for transferability estimation. Our empirical results of using task-relatedness to select the best pre-trained model from a model zoo for a target task highlight its utility for practical problems.