Deep Least Squares Alignment for Unsupervised Domain Adaptation

TL;DR

This paper introduces Deep Least Squares Alignment (DLSA), a novel method for unsupervised domain adaptation that aligns domain distributions in a latent space using a linear model, improving performance over existing techniques.

Contribution

The paper proposes DLSA, a new approach that estimates domain distributions via a linear model and introduces marginal and conditional adaptation losses for better alignment.

Findings

DLSA effectively aligns domain distributions in experiments.

DLSA outperforms state-of-the-art methods in unsupervised domain adaptation.

The method reduces domain discrepancy by minimizing angles and intercept differences.

Abstract

Unsupervised domain adaptation leverages rich information from a labeled source domain to model an unlabeled target domain. Existing methods attempt to align the cross-domain distributions. However, the statistical representations of the alignment of the two domains are not well addressed. In this paper, we propose deep least squares alignment (DLSA) to estimate the distribution of the two domains in a latent space by parameterizing a linear model. We further develop marginal and conditional adaptation loss to reduce the domain discrepancy by minimizing the angle between fitting lines and intercept differences and further learning domain invariant features. Extensive experiments demonstrate that the proposed DLSA model is effective in aligning domain distributions and outperforms state-of-the-art methods.