Practical Transferability Estimation for Image Classification Tasks

TL;DR

This paper introduces JC-NCE, a practical transferability metric for image classification that improves robustness and accuracy without needing auxiliary tasks, outperforming existing methods in various transfer settings.

Contribution

The paper proposes JC-NCE, a new transferability score that enhances robustness and efficiency by eliminating auxiliary tasks and leveraging optimal transport for better task difference estimation.

Findings

JC-NCE outperforms OTCE by 7-12% in transferability accuracy.

JC-NCE is more robust than existing metrics across datasets.

The method reduces computational overhead compared to prior approaches.

Abstract

Transferability estimation is an essential problem in transfer learning to predict how good the performance is when transferring a source model (or source task) to a target task. Recent analytical transferability metrics have been widely used for source model selection and multi-task learning. A major challenge is how to make transfereability estimation robust under the cross-domain cross-task settings. The recently proposed OTCE score solves this problem by considering both domain and task differences, with the help of transfer experiences on auxiliary tasks, which causes an efficiency overhead. In this work, we propose a practical transferability metric called JC-NCE score that dramatically improves the robustness of the task difference estimation in OTCE, thus removing the need for auxiliary tasks. Specifically, we build the joint correspondences between source and target data via solving an optimal transport problem with a ground cost considering both the sample distance and label distance, and then compute the transferability score as the negative conditional entropy of the matched labels. Extensive validations under the intra-dataset and inter-dataset transfer settings demonstrate that our JC-NCE score outperforms the auxiliary-task free version of OTCE for 7% and 12%, respectively, and is also more robust than other existing transferability metrics on average.