Transfer Learning with Dynamic Distribution Adaptation

TL;DR

This paper introduces Dynamic Distribution Adaptation (DDA), a method that quantitatively evaluates and leverages the different contributions of marginal and conditional distributions to improve transfer learning across various tasks.

Contribution

The paper proposes DDA, a novel approach to quantify and utilize the importance of distribution types, along with two algorithms MDDA and DDAN for transfer learning.

Findings

DDA effectively evaluates the importance of distribution types.

MDDA and DDAN outperform existing transfer learning methods.

The approach improves performance on tasks like digit recognition, sentiment analysis, and image classification.

Abstract

Transfer learning aims to learn robust classifiers for the target domain by leveraging knowledge from a source domain. Since the source and the target domains are usually from different distributions, existing methods mainly focus on adapting the cross-domain marginal or conditional distributions. However, in real applications, the marginal and conditional distributions usually have different contributions to the domain discrepancy. Existing methods fail to quantitatively evaluate the different importance of these two distributions, which will result in unsatisfactory transfer performance. In this paper, we propose a novel concept called Dynamic Distribution Adaptation (DDA), which is capable of quantitatively evaluating the relative importance of each distribution. DDA can be easily incorporated into the framework of structural risk minimization to solve transfer learning problems. On the basis of DDA, we propose two novel learning algorithms: (1) Manifold Dynamic Distribution Adaptation (MDDA) for traditional transfer learning, and (2) Dynamic Distribution Adaptation Network (DDAN) for deep transfer learning. Extensive experiments demonstrate that MDDA and DDAN significantly improve the transfer learning performance and setup a strong baseline over the latest deep and adversarial methods on digits recognition, sentiment analysis, and image classification. More importantly, it is shown that marginal and conditional distributions have different contributions to the domain divergence, and our DDA is able to provide good quantitative evaluation of their relative importance which leads to better performance. We believe this observation can be helpful for future research in transfer learning.