This repository contains a demo of our algorithm, Dataset Dictionary Learning (DaDiL). This consists on the toy example used throughout the paper.
Our paper was accepted on the 26th European Conference on Artificial Intelligence (ECAI 2023). You can check our pre-print on ArXiv
This paper seeks to solve Multi-Source Domain Adaptation (MSDA), which aims to mitigate data distribution shifts when transferring knowledge from multiple labeled source domains to an unlabeled target domain. We propose a novel MSDA framework based on dictionary learning and optimal transport. We interpret each domain in MSDA as an empirical distribution. As such, we express each domain as a Wasserstein barycenter of dictionary atoms, which are empirical distributions. We propose a novel algorithm, DaDiL, for learning via mini-batches: (i) atom distributions; (ii) a matrix of barycentric coordinates. Based on our dictionary, we propose two novel methods for MSDA: DaDil-R, based on the reconstruction of labeled samples in the target domain, and DaDiL-E, based on the ensembling of classifiers learned on atom distributions. We evaluate our methods in 3 benchmarks: Caltech-Office, Refurbished Office 31, and CRWU, where we improved previous state-of-the-art by 3.15%, 2.29%, and 7.71% in classification performance. Finally, we show that interpolations in the Wasserstein hull of learned atoms provide data that can generalize to the target domain
We experimented on 3 Multi-Source Domain Adaptation benchmarks: (i) Caltech-Office 10, (ii) Refurbished Office 31 and (iii) CWRU. The results are shown in the tables below,
@article{montesuma2023multi,
title={Multi-Source Domain Adaptation through Dataset Dictionary Learning in Wasserstein Space},
author={Montesuma, Eduardo Fernandes and Mboula, Fred Ngol{\`e} and Souloumiac, Antoine},
journal={arXiv preprint arXiv:2307.14953},
year={2023}
}