Robust Class-Conditional Distribution Alignment for Partial Domain Adaptation

TL;DR

This paper introduces a robust class-conditional distribution alignment method for partial domain adaptation, addressing negative transfer by focusing on class-level feature alignment and robust pseudo-labeling.

Contribution

It proposes a novel approach that goes beyond first-order moments, optimizing intra- and inter-class distributions with a complement entropy module for improved domain adaptation.

Findings

Outperforms benchmark methods in partial domain adaptation tasks.

Effective in reducing negative transfer and classification uncertainty.

Ablation studies confirm the contribution of each module.

Abstract

Unwanted samples from private source categories in the learning objective of a partial domain adaptation setup can lead to negative transfer and reduce classification performance. Existing methods, such as re-weighting or aggregating target predictions, are vulnerable to this issue, especially during initial training stages, and do not adequately address class-level feature alignment. Our proposed approach seeks to overcome these limitations by delving deeper than just the first-order moments to derive distinct and compact categorical distributions. We employ objectives that optimize the intra and inter-class distributions in a domain-invariant fashion and design a robust pseudo-labeling for efficient target supervision. Our approach incorporates a complement entropy objective module to reduce classification uncertainty and flatten incorrect category predictions. The experimental findings and ablation analysis of the proposed modules demonstrate the superior performance of our proposed model compared to benchmarks.