Transferability Estimation Based On Principal Gradient Expectation

TL;DR

This paper introduces Principal Gradient Expectation (PGE), a novel transferability estimation method that evaluates source-target domain suitability by analyzing the expectation of gradients, balancing performance and computational cost.

Contribution

The paper proposes PGE, a new transferability estimation technique based on principal gradient expectation, improving stability, reliability, and efficiency over existing methods.

Findings

PGE outperforms state-of-the-art transferability metrics in experiments.

PGE effectively balances estimation accuracy and computational cost.

Extensive tests confirm PGE's superiority across multiple properties.

Abstract

Transfer learning aims to improve the performance of target tasks by transferring knowledge acquired in source tasks. The standard approach is pre-training followed by fine-tuning or linear probing. Especially, selecting a proper source domain for a specific target domain under predefined tasks is crucial for improving efficiency and effectiveness. It is conventional to solve this problem via estimating transferability. However, existing methods can not reach a trade-off between performance and cost. To comprehensively evaluate estimation methods, we summarize three properties: stability, reliability and efficiency. Building upon them, we propose Principal Gradient Expectation(PGE), a simple yet effective method for assessing transferability. Specifically, we calculate the gradient over each weight unit multiple times with a restart scheme, and then we compute the expectation of all gradients. Finally, the transferability between the source and target is estimated by computing the gap of normalized principal gradients. Extensive experiments show that the proposed metric is superior to state-of-the-art methods on all properties.