Interpretable Domain Adaptation Using ...
Document type :
Compte-rendu et recension critique d'ouvrage
Title :
Interpretable Domain Adaptation Using Unsupervised Feature Selection on Pre-trained Source Models
Author(s) :
Zhang, Luxin [Auteur]
MOdel for Data Analysis and Learning [MODAL]
Worldline France
Germain, Pascal [Auteur]
Université Laval [Québec] [ULaval]
Kessaci, Yacine [Auteur]
Worldline France
Biernacki, Christophe [Auteur]
MOdel for Data Analysis and Learning [MODAL]
MOdel for Data Analysis and Learning [MODAL]
Worldline France
Germain, Pascal [Auteur]
Université Laval [Québec] [ULaval]
Kessaci, Yacine [Auteur]
Worldline France
Biernacki, Christophe [Auteur]
MOdel for Data Analysis and Learning [MODAL]
Journal title :
Neurocomputing
Pages :
319-336
Publisher :
Elsevier
Publication date :
2022-10-28
ISSN :
0925-2312
HAL domain(s) :
Informatique [cs]/Apprentissage [cs.LG]
English abstract : [en]
We study a realistic domain adaptation setting where one has access to an already existing "black-box" machine learning model. Indeed, in real-life scenarios, an efficient pre-trained source domain predictive model is often ...
Show more >We study a realistic domain adaptation setting where one has access to an already existing "black-box" machine learning model. Indeed, in real-life scenarios, an efficient pre-trained source domain predictive model is often available and required to be preserved. Our work extends a method that has been recently proposed to tackle this specific problem, yet providing an interpretable <i>target to source</i> transformation, by seeking a coordinate-wise adaptation of the feature space. However, this method requires partially labeled target data to select the features to be adapted. In contrast, we address the more challenging unsupervised version of this domain adaptation scenario. We propose a new pseudo-label estimator over unlabeled target examples, based on the rank-stability in regards to the source model prediction. Such estimated "labels" are further used in a feature selection process to assess whether each feature needs to be transformed to achieve adaptation. We provide theoretical foundations of our method as well as an efficient implementation. Numerical experiments on real datasets show particularly encouraging results since approaching the supervised case, where one has access to labeled target samples.Show less >
Show more >We study a realistic domain adaptation setting where one has access to an already existing "black-box" machine learning model. Indeed, in real-life scenarios, an efficient pre-trained source domain predictive model is often available and required to be preserved. Our work extends a method that has been recently proposed to tackle this specific problem, yet providing an interpretable <i>target to source</i> transformation, by seeking a coordinate-wise adaptation of the feature space. However, this method requires partially labeled target data to select the features to be adapted. In contrast, we address the more challenging unsupervised version of this domain adaptation scenario. We propose a new pseudo-label estimator over unlabeled target examples, based on the rank-stability in regards to the source model prediction. Such estimated "labels" are further used in a feature selection process to assess whether each feature needs to be transformed to achieve adaptation. We provide theoretical foundations of our method as well as an efficient implementation. Numerical experiments on real datasets show particularly encouraging results since approaching the supervised case, where one has access to labeled target samples.Show less >
Language :
Anglais
Popular science :
Non
Collections :
Source :
Files
- document
- Open access
- Access the document
- preprint.pdf
- Open access
- Access the document