Multi-Source Domain Adaptation through Dataset Dictionary Learning in Wasserstein Space

Antoine Souloumiac; Eduardo Fernandes Montesuma; Fred Ngol\`e Mboula

RECORD

Multi-Source Domain Adaptation through Dataset Dictionary Learning in Wasserstein Space

Eduardo Fernandes Montesuma, Fred Ngolè Mboula, Antoine Souloumiac

Machine Learning · Submitted 27 Jul 2023

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim We evaluate our methods in 3 benchmarks: Caltech-Office, Office 31, and CRWU, where we improved previous state-of-the-art by 3.15%, 2.29%, and 7.71% in classification performance.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2307.14953 (version 3)

Pith IDFTPZEVWA, Pith's permanent record of this paper and its review

CitationAntoine Souloumiac, Eduardo Fernandes Montesuma, Fred Ngol\`e Mboula. "Multi-Source Domain Adaptation through Dataset Dictionary Learning in Wasserstein Space." arXiv:2307.14953 (2023). https://pith.science/paper/FTPZEVWA

Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 6,309 input tokens.

What the abstract itself states, as probabilities: claims first 0.11 · states limitation 0.06 · human participants 0.04 · controlled comparison 0.52 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.96 · mentions code or data availability 0.03.

The abstract, sentence by sentence

arXiv listing abstract v3. Results and the main claim are underlined; show every role.

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, 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.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	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.	background 1.00	0.21	0.04	asserted 0.92
s1	We propose a novel MSDA framework based on dictionary learning and optimal transport.	contribution 0.94	0.91	0.07	asserted 0.99
s2	We interpret each domain in MSDA as an empirical distribution.	method 0.74	0.51	0.02	asserted 1.00
s3	As such, we express each domain as a Wasserstein barycenter of dictionary atoms, which are empirical distributions.	method 0.80	0.83	0.06	asserted 1.00
s4	We propose a novel algorithm, DaDiL, for learning via mini-batches: (i) atom distributions; (ii) a matrix of barycentric coordinates.	method 0.73	0.95	0.04	asserted 1.00
s5	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.	method 0.56	0.93	0.09	asserted 0.98
s6	We evaluate our methods in 3 benchmarks: Caltech-Office, Office 31, and CRWU, where we improved previous state-of-the-art by 3.15%, 2.29%, and 7.71% in classification performance.	result 1.00	0.97	0.27	asserted 1.00
s7	Finally, we show that interpolations in the Wasserstein hull of learned atoms provide data that can generalize to the target domain.	result 0.84	0.95	0.38	asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.71.

What was examined, and what was not

sources read

arXiv listing abstract v3

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper

Original abstract

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, 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.

Human review

How Pith checks papers → · Full audit of this paper →

How the literature uses it

2 arXiv papers cite this in 6 sentences

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

Papers citing it · 2 sentences without a confident label

Compares with it 1 paper

We also compare two multi-source domain adaptation methods [ref] and [ref] .

Multiply Robust Estimation for Local Distribution Shifts with Multiple Domains · Experiments · 1.00

Cites it as support 1 paper

These remarks agree with previous works in da [ref] , and cdfd [ref] , and show the power of ot as a framework for both ssda and msda.

Benchmarking Domain Adaptation for Chemical Processes on the Tennessee Eastman Process · Multi-Source CDFD · 0.96

Cites it as background 2 papers

Usually, msda methods rely on moment matching [ref] , domain re-weighting [ref] , or distribution matching [ref] .

Benchmarking Domain Adaptation for Chemical Processes on the Tennessee Eastman Process · Introduction · 0.88
Another research direction leverages optimal transport to learn a dictionary of data distributions and then approximate the target population using a mixture of the elements in the dictionary [ref] .

Multiply Robust Estimation for Local Distribution Shifts with Multiple Domains · Related Work · 0.93

What it cites · 2 sentences without a confident label

Uses the method or tool of 1 paper

We thus employ mini-batch ot [ref] .

Minibatch optimal transport distances; analysis and applications · Dataset Dictionary Learning for MSDA · 1.00

Cites as background 1 paper

We refer readers to [ref] and [ref] for further background on this topic.

Recent Advances in Optimal Transport for Machine Learning · Optimal Transport · 1.00

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.