Exploring Model Transferability through the Lens of Potential Energy

TL;DR

This paper introduces PED, a physics-inspired method that models transferability of pre-trained models through potential energy dynamics, improving model selection for downstream tasks.

Contribution

It proposes a novel physics-based approach to evaluate transferability by modeling representation dynamics as forces affecting potential energy, enhancing stability and reliability.

Findings

Improves transferability estimation accuracy across multiple tasks

Seamlessly integrates with existing ranking methods

Demonstrates effectiveness on 10 downstream tasks and 12 models

Abstract

Transfer learning has become crucial in computer vision tasks due to the vast availability of pre-trained deep learning models. However, selecting the optimal pre-trained model from a diverse pool for a specific downstream task remains a challenge. Existing methods for measuring the transferability of pre-trained models rely on statistical correlations between encoded static features and task labels, but they overlook the impact of underlying representation dynamics during fine-tuning, leading to unreliable results, especially for self-supervised models. In this paper, we present an insightful physics-inspired approach named PED to address these challenges. We reframe the challenge of model selection through the lens of potential energy and directly model the interaction forces that influence fine-tuning dynamics. By capturing the motion of dynamic representations to decline the potential energy within a force-driven physical model, we can acquire an enhanced and more stable observation for estimating transferability. The experimental results on 10 downstream tasks and 12 self-supervised models demonstrate that our approach can seamlessly integrate into existing ranking techniques and enhance their performances, revealing its effectiveness for the model selection task and its potential for understanding the mechanism in transfer learning. Code will be available at https://github.com/lixiaotong97/PED.