21.
Ruf, Boris; Detyniecki, Marcin
Towards the right kind of fairness in AI Journal Article
In: arXiv preprint arXiv:2102.08453, 2021.
@article{ruf2021towards,
title = {Towards the right kind of fairness in AI},
author = {Boris Ruf and Marcin Detyniecki},
year = {2021},
date = {2021-01-01},
journal = {arXiv preprint arXiv:2102.08453},
keywords = {Fairness},
pubstate = {published},
tppubtype = {article}
}
22.
Ruf, Boris; Detyniecki, Marcin
Implementing Fair Regression In The Real World Journal Article
In: arXiv preprint arXiv:2104.04353, 2021.
@article{ruf2021implementing,
title = {Implementing Fair Regression In The Real World},
author = {Boris Ruf and Marcin Detyniecki},
year = {2021},
date = {2021-01-01},
journal = {arXiv preprint arXiv:2104.04353},
keywords = {Fairness},
pubstate = {published},
tppubtype = {article}
}
23.
Bove, Clara; Aigrain, Jonathan; Detyniecki, Marcin
Building Trust in Artificial Conversational Agents Journal Article
In: IUI 2021 Workshop on Conversational User Interface (CUI), 2021.
BibTeX | Tags: Human-Computer Interface & HCXAI
@article{bove2021trust,
title = {Building Trust in Artificial Conversational
Agents},
author = {Clara Bove and Jonathan Aigrain and Marcin Detyniecki},
year = {2021},
date = {2021-01-01},
journal = {IUI 2021 Workshop on Conversational User Interface (CUI)},
keywords = {Human-Computer Interface & HCXAI},
pubstate = {published},
tppubtype = {article}
}
24.
Bove, Clara; Aigrain, Jonathan; Lesot, Marie-Jeanne; Tijus, Charles; Detyniecki, Marcin
Contextualization and Exploration of Local Feature Importance Explanations to Improve Understanding and Satisfaction of Non-Expert Users Journal Article
In: IUI 2021 Workshop Transparent and Explainable Smart System (TExSS), 2021.
BibTeX | Tags: Human-Computer Interface & HCXAI
@article{bove2021contextualization,
title = {Contextualization and Exploration of Local Feature Importance Explanations to Improve Understanding and Satisfaction of Non-Expert Users},
author = {Clara Bove and Jonathan Aigrain and Marie-Jeanne Lesot and Charles Tijus and Marcin Detyniecki},
year = {2021},
date = {2021-01-01},
journal = {IUI 2021 Workshop Transparent and Explainable Smart System (TExSS)},
keywords = {Human-Computer Interface & HCXAI},
pubstate = {published},
tppubtype = {article}
}
25.
Rychener, Yves; Renard, Xavier; Seddah, Djamé; Frossard, Pascal; Detyniecki, Marcin
Quackie: A nlp classification task with ground truth explanations Journal Article
In: arXiv preprint arXiv:2012.13190, 2020.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@article{rychener2020quackie,
title = {Quackie: A nlp classification task with ground truth explanations},
author = {Yves Rychener and Xavier Renard and Djamé Seddah and Pascal Frossard and Marcin Detyniecki},
url = {https://arxiv.org/abs/2012.13190},
year = {2020},
date = {2020-01-01},
journal = {arXiv preprint arXiv:2012.13190},
abstract = {NLP Interpretability aims to increase trust in model predictions. This makes evaluating interpretability approaches a pressing issue. There are multiple datasets for evaluating NLP Interpretability, but their dependence on human provided ground truths raises questions about their unbiasedness. In this work, we take a different approach and formulate a specific classification task by diverting question-answering datasets. For this custom classification task, the interpretability ground-truth arises directly from the definition of the classification problem. We use this method to propose a benchmark and lay the groundwork for future research in NLP interpretability by evaluating a wide range of current state of the art methods.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {article}
}
NLP Interpretability aims to increase trust in model predictions. This makes evaluating interpretability approaches a pressing issue. There are multiple datasets for evaluating NLP Interpretability, but their dependence on human provided ground truths raises questions about their unbiasedness. In this work, we take a different approach and formulate a specific classification task by diverting question-answering datasets. For this custom classification task, the interpretability ground-truth arises directly from the definition of the classification problem. We use this method to propose a benchmark and lay the groundwork for future research in NLP interpretability by evaluating a wide range of current state of the art methods.
26.
Laugel, Thibault
Local Post-hoc Interpretability for Black-box Classifiers PhD Thesis
Sorbonne Université, CNRS, LIP6, F-75005 Paris, France, 2020.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@phdthesis{laugel2020local,
title = {Local Post-hoc Interpretability for Black-box Classifiers},
author = {Thibault Laugel},
url = {https://hal.archives-ouvertes.fr/tel-03002496v1},
year = {2020},
date = {2020-01-01},
school = {Sorbonne Université, CNRS, LIP6, F-75005 Paris, France},
abstract = {This thesis focuses on the field of XAI (eXplainable AI), and more particularly local post-hoc interpretability paradigm, that is to say the generation of explanations for a single prediction of a trained classifier. In particular, we study a fully agnostic context, meaning that the explanation is generated without using any knowledge about the classifier (treated as a black-box) nor the data used to train it. In this thesis, we identify several issues that can arise in this context and that may be harmful for interpretability. We propose to study each of these issues and propose novel criteria and approaches to detect and characterize them. The three issues we focus on are: the risk of generating explanations that are out of distribution; the risk of generating explanations that cannot be associated to any ground-truth instance; and the risk of generating explanations that are not local enough. These risks are studied through two specific categories of interpretability approaches: counterfactual explanations, and local surrogate models.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {phdthesis}
}
This thesis focuses on the field of XAI (eXplainable AI), and more particularly local post-hoc interpretability paradigm, that is to say the generation of explanations for a single prediction of a trained classifier. In particular, we study a fully agnostic context, meaning that the explanation is generated without using any knowledge about the classifier (treated as a black-box) nor the data used to train it. In this thesis, we identify several issues that can arise in this context and that may be harmful for interpretability. We propose to study each of these issues and propose novel criteria and approaches to detect and characterize them. The three issues we focus on are: the risk of generating explanations that are out of distribution; the risk of generating explanations that cannot be associated to any ground-truth instance; and the risk of generating explanations that are not local enough. These risks are studied through two specific categories of interpretability approaches: counterfactual explanations, and local surrogate models.
27.
Grari, Vincent; Ruf, Boris; Lamprier, Sylvain; Detyniecki, Marcin
Achieving fairness with decision trees: An adversarial approach Journal Article
In: Data Science and Engineering, vol. 5, no. 2, pp. 99–110, 2020.
@article{grari2020achieving,
title = {Achieving fairness with decision trees: An adversarial approach},
author = {Vincent Grari and Boris Ruf and Sylvain Lamprier and Marcin Detyniecki},
year = {2020},
date = {2020-01-01},
journal = {Data Science and Engineering},
volume = {5},
number = {2},
pages = {99--110},
publisher = {Springer},
keywords = {Fairness},
pubstate = {published},
tppubtype = {article}
}
28.
Ruf, Boris; Boutharouite, Chaouki; Detyniecki, Marcin
Getting Fairness Right: Towards a Toolbox for Practitioners Journal Article
In: arXiv preprint arXiv:2003.06920, 2020.
@article{ruf2020getting,
title = {Getting Fairness Right: Towards a Toolbox for Practitioners},
author = {Boris Ruf and Chaouki Boutharouite and Marcin Detyniecki},
year = {2020},
date = {2020-01-01},
journal = {arXiv preprint arXiv:2003.06920},
keywords = {Fairness},
pubstate = {published},
tppubtype = {article}
}
29.
Renard, Xavier; Woloszko, Nicolas; Aigrain, Jonathan; Detyniecki, Marcin
Concept tree: High-level representation of variables for more interpretable surrogate decision trees Journal Article
In: Bank of England and King's College London joint conference on Modelling with Big Data and Machine Learning: Interpretability and Model Uncertainty, 2019.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@article{renard2019concept,
title = {Concept tree: High-level representation of variables for more interpretable surrogate decision trees},
author = {Xavier Renard and Nicolas Woloszko and Jonathan Aigrain and Marcin Detyniecki},
url = {https://arxiv.org/abs/1906.01297},
year = {2019},
date = {2019-01-01},
journal = {Bank of England and King's College London joint conference on Modelling with Big Data and Machine Learning: Interpretability and Model Uncertainty},
abstract = {Interpretable surrogates of black-box predictors trained on high-dimensional tabular datasets can struggle to generate comprehensible explanations in the presence of correlated variables. We propose a model-agnostic interpretable surrogate that provides global and local explanations of black-box classifiers to address this issue. We introduce the idea of concepts as intuitive groupings of variables that are either defined by a domain expert or automatically discovered using correlation coefficients. Concepts are embedded in a surrogate decision tree to enhance its comprehensibility. First experiments on FRED-MD, a macroeconomic database with 134 variables, show improvement in human-interpretability while accuracy and fidelity of the surrogate model are preserved.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {article}
}
Interpretable surrogates of black-box predictors trained on high-dimensional tabular datasets can struggle to generate comprehensible explanations in the presence of correlated variables. We propose a model-agnostic interpretable surrogate that provides global and local explanations of black-box classifiers to address this issue. We introduce the idea of concepts as intuitive groupings of variables that are either defined by a domain expert or automatically discovered using correlation coefficients. Concepts are embedded in a surrogate decision tree to enhance its comprehensibility. First experiments on FRED-MD, a macroeconomic database with 134 variables, show improvement in human-interpretability while accuracy and fidelity of the surrogate model are preserved.
30.
Renard, Xavier; Woloszko, Nicolas; Aigrain, Jonathan; Detyniecki, Marcin
Concept tree: High-level representation of variables for more interpretable surrogate decision trees Journal Article
In: International Conference on Machine Learning (ICML) Workshop on Human In the Loop Learning (HILL), 2019.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@article{renard2019conceptb,
title = {Concept tree: High-level representation of variables for more interpretable surrogate decision trees},
author = {Xavier Renard and Nicolas Woloszko and Jonathan Aigrain and Marcin Detyniecki},
url = {https://arxiv.org/abs/1906.01297},
year = {2019},
date = {2019-01-01},
journal = {International Conference on Machine Learning (ICML) Workshop on Human In the Loop Learning (HILL)},
abstract = {Interpretable surrogates of black-box predictors trained on high-dimensional tabular datasets can struggle to generate comprehensible explanations in the presence of correlated variables. We propose a model-agnostic interpretable surrogate that provides global and local explanations of black-box classifiers to address this issue. We introduce the idea of concepts as intuitive groupings of variables that are either defined by a domain expert or automatically discovered using correlation coefficients. Concepts are embedded in a surrogate decision tree to enhance its comprehensibility. First experiments on FRED-MD, a macroeconomic database with 134 variables, show improvement in human-interpretability while accuracy and fidelity of the surrogate model are preserved.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {article}
}
Interpretable surrogates of black-box predictors trained on high-dimensional tabular datasets can struggle to generate comprehensible explanations in the presence of correlated variables. We propose a model-agnostic interpretable surrogate that provides global and local explanations of black-box classifiers to address this issue. We introduce the idea of concepts as intuitive groupings of variables that are either defined by a domain expert or automatically discovered using correlation coefficients. Concepts are embedded in a surrogate decision tree to enhance its comprehensibility. First experiments on FRED-MD, a macroeconomic database with 134 variables, show improvement in human-interpretability while accuracy and fidelity of the surrogate model are preserved.
31.
Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Renard, Xavier; Detyniecki, Marcin
The Dangers of Post-hoc Interpretability: Unjustified Counterfactual Explanations Proceedings Article
In: Twenty-Eighth International Joint Conference on Artificial Intelligence $$IJCAI-19$$, pp. 2801–2807, International Joint Conferences on Artificial Intelligence Organization 2019.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@inproceedings{laugel2019dangers,
title = {The Dangers of Post-hoc Interpretability: Unjustified Counterfactual Explanations},
author = {Thibault Laugel and Marie-Jeanne Lesot and Christophe Marsala and Xavier Renard and Marcin Detyniecki},
url = {https://arxiv.org/abs/1907.09294},
year = {2019},
date = {2019-01-01},
booktitle = {Twenty-Eighth International Joint Conference on Artificial Intelligence $$IJCAI-19$$},
pages = {2801--2807},
organization = {International Joint Conferences on Artificial Intelligence Organization},
abstract = {Post-hoc interpretability approaches have been proven to be powerful tools to generate explanations for the predictions made by a trained black-box model. However, they create the risk of having explanations that are a result of some artifacts learned by the model instead of actual knowledge from the data. This paper focuses on the case of counterfactual explanations and asks whether the generated instances can be justified, i.e. continuously connected to some ground-truth data. We evaluate the risk of generating unjustified counterfactual examples by investigating the local neighborhoods of instances whose predictions are to be explained and show that this risk is quite high for several datasets. Furthermore, we show that most state of the art approaches do not differentiate justified from unjustified counterfactual examples, leading to less useful explanations.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {inproceedings}
}
Post-hoc interpretability approaches have been proven to be powerful tools to generate explanations for the predictions made by a trained black-box model. However, they create the risk of having explanations that are a result of some artifacts learned by the model instead of actual knowledge from the data. This paper focuses on the case of counterfactual explanations and asks whether the generated instances can be justified, i.e. continuously connected to some ground-truth data. We evaluate the risk of generating unjustified counterfactual examples by investigating the local neighborhoods of instances whose predictions are to be explained and show that this risk is quite high for several datasets. Furthermore, we show that most state of the art approaches do not differentiate justified from unjustified counterfactual examples, leading to less useful explanations.
32.
Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Renard, Xavier; Detyniecki, Marcin
Unjustified classification regions and counterfactual explanations in machine learning Proceedings Article
In: Joint European conference on machine learning and knowledge discovery in databases, pp. 37–54, Springer 2019.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@inproceedings{laugel2019unjustified,
title = {Unjustified classification regions and counterfactual explanations in machine learning},
author = {Thibault Laugel and Marie-Jeanne Lesot and Christophe Marsala and Xavier Renard and Marcin Detyniecki},
url = {https://ecmlpkdd2019.org/downloads/paper/226.pdf},
year = {2019},
date = {2019-01-01},
booktitle = {Joint European conference on machine learning and knowledge discovery in databases},
pages = {37--54},
organization = {Springer},
abstract = {Post-hoc interpretability approaches, although powerful tools to generate explanations for predictions made by a trained black-box model, have been shown to be vulnerable to issues caused by lack of robustness of the classifier. In particular, this paper focuses on the notion of explanation justification, defined as connectedness to ground-truth data, in the context of counterfactuals. In this work, we explore the extent of the risk of generating unjustified explanations. We propose an empirical study to assess the vulnerability of classifiers and show that the chosen learning algorithm heavily impacts the vulnerability of the model. Additionally, we show that state-of-the-art post-hoc counterfactual approaches can minimize the impact of this risk by generating less local explanations.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {inproceedings}
}
Post-hoc interpretability approaches, although powerful tools to generate explanations for predictions made by a trained black-box model, have been shown to be vulnerable to issues caused by lack of robustness of the classifier. In particular, this paper focuses on the notion of explanation justification, defined as connectedness to ground-truth data, in the context of counterfactuals. In this work, we explore the extent of the risk of generating unjustified explanations. We propose an empirical study to assess the vulnerability of classifiers and show that the chosen learning algorithm heavily impacts the vulnerability of the model. Additionally, we show that state-of-the-art post-hoc counterfactual approaches can minimize the impact of this risk by generating less local explanations.
33.
Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Detyniecki, Marcin
Issues with post-hoc counterfactual explanations: a discussion Proceedings Article
In: ICML Workshop on Human in the Loop Learning (HILL 2019), 2019.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@inproceedings{laugel2019issues,
title = {Issues with post-hoc counterfactual explanations: a discussion},
author = {Thibault Laugel and Marie-Jeanne Lesot and Christophe Marsala and Marcin Detyniecki},
url = {https://arxiv.org/abs/1906.04774},
year = {2019},
date = {2019-01-01},
booktitle = {ICML Workshop on Human in the Loop Learning (HILL 2019)},
abstract = {Counterfactual post-hoc interpretability approaches have been proven to be useful tools to generate explanations for the predictions of a trained blackbox classifier. However, the assumptions they make about the data and the classifier make them unreliable in many contexts. In this paper, we discuss three desirable properties and approaches to quantify them: proximity, connectedness and stability. In addition, we illustrate that there is a risk for post-hoc counterfactual approaches to not satisfy these properties.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {inproceedings}
}
Counterfactual post-hoc interpretability approaches have been proven to be useful tools to generate explanations for the predictions of a trained blackbox classifier. However, the assumptions they make about the data and the classifier make them unreliable in many contexts. In this paper, we discuss three desirable properties and approaches to quantify them: proximity, connectedness and stability. In addition, we illustrate that there is a risk for post-hoc counterfactual approaches to not satisfy these properties.
34.
Grari, Vincent; Ruf, Boris; Lamprier, Sylvain; Detyniecki, Marcin
Fair adversarial gradient tree boosting Proceedings Article
In: 2019 IEEE International Conference on Data Mining (ICDM), pp. 1060–1065, IEEE 2019.
@inproceedings{grari2019fair,
title = {Fair adversarial gradient tree boosting},
author = {Vincent Grari and Boris Ruf and Sylvain Lamprier and Marcin Detyniecki},
year = {2019},
date = {2019-01-01},
booktitle = {2019 IEEE International Conference on Data Mining (ICDM)},
pages = {1060--1065},
organization = {IEEE},
keywords = {Fairness},
pubstate = {published},
tppubtype = {inproceedings}
}
35.
Ballet, Vincent; Renard, Xavier; Aigrain, Jonathan; Laugel, Thibault; Frossard, Pascal; Detyniecki, Marcin
Imperceptible adversarial attacks on tabular data Journal Article
In: NeurIPS 2019 Workshop on Robust AI in Financial Services: Data, Fairness, Explainability, Trustworthiness, and Privacy (Robust AI in FS 2019), Vancouver, Canada, 2019.
Abstract | Links | BibTeX | Tags: Robustness
@article{ballet2019imperceptible,
title = {Imperceptible adversarial attacks on tabular data},
author = {Vincent Ballet and Xavier Renard and Jonathan Aigrain and Thibault Laugel and Pascal Frossard and Marcin Detyniecki},
url = {https://arxiv.org/abs/1911.03274},
year = {2019},
date = {2019-01-01},
journal = {NeurIPS 2019 Workshop on Robust AI in Financial Services: Data, Fairness, Explainability, Trustworthiness, and Privacy (Robust AI in FS 2019), Vancouver, Canada},
abstract = {Security of machine learning models is a concern as they may face adversarial attacks for unwarranted advantageous decisions. While research on the topic has mainly been focusing on the image domain, numerous industrial applications, in particular in finance, rely on standard tabular data. In this paper, we discuss the notion of adversarial examples in the tabular domain. We propose a formalization based on the imperceptibility of attacks in the tabular domain leading to an approach to generate imperceptible adversarial examples. Experiments show that we can generate imperceptible adversarial examples with a high fooling rate.},
keywords = {Robustness},
pubstate = {published},
tppubtype = {article}
}
Security of machine learning models is a concern as they may face adversarial attacks for unwarranted advantageous decisions. While research on the topic has mainly been focusing on the image domain, numerous industrial applications, in particular in finance, rely on standard tabular data. In this paper, we discuss the notion of adversarial examples in the tabular domain. We propose a formalization based on the imperceptibility of attacks in the tabular domain leading to an approach to generate imperceptible adversarial examples. Experiments show that we can generate imperceptible adversarial examples with a high fooling rate.
36.
Aigrain, Jonathan; Detyniecki, Marcin
Detecting adversarial examples and other misclassifications in neural networks by introspection Journal Article
In: arXiv preprint arXiv:1905.09186, 2019.
BibTeX | Tags: Robustness
@article{aigrain2019detecting,
title = {Detecting adversarial examples and other misclassifications in neural networks by introspection},
author = {Jonathan Aigrain and Marcin Detyniecki},
year = {2019},
date = {2019-01-01},
journal = {arXiv preprint arXiv:1905.09186},
keywords = {Robustness},
pubstate = {published},
tppubtype = {article}
}
37.
Renard, Xavier; Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Detyniecki, Marcin
Detecting potential local adversarial examples for human-interpretable defense Proceedings Article
In: ECML PKDD 2018 Workshops: Nemesis 2018, UrbReas 2018, SoGood 2018, IWAISe 2018, and Green Data Mining 2018, Dublin, Ireland, September 10-14, 2018, Proceedings, pp. 41, Springer 2019.
Abstract | Links | BibTeX | Tags: Robustness
@inproceedings{renard2018detecting,
title = {Detecting potential local adversarial examples for human-interpretable defense},
author = {Xavier Renard and Thibault Laugel and Marie-Jeanne Lesot and Christophe Marsala and Marcin Detyniecki},
url = {https://arxiv.org/abs/1809.02397},
year = {2019},
date = {2019-01-01},
booktitle = {ECML PKDD 2018 Workshops: Nemesis 2018, UrbReas 2018, SoGood 2018, IWAISe 2018, and Green Data Mining 2018, Dublin, Ireland, September 10-14, 2018, Proceedings},
volume = {11329},
pages = {41},
organization = {Springer},
abstract = {Machine learning models are increasingly used in the industry to make decisions such as credit insurance approval. Some people may be tempted to manipulate specific variables, such as the age or the salary, in order to get better chances of approval. In this ongoing work, we propose to discuss, with a first proposition, the issue of detecting a potential local adversarial example on classical tabular data by providing to a human expert the locally critical features for the classifier's decision, in order to control the provided information and avoid a fraud.},
keywords = {Robustness},
pubstate = {published},
tppubtype = {inproceedings}
}
Machine learning models are increasingly used in the industry to make decisions such as credit insurance approval. Some people may be tempted to manipulate specific variables, such as the age or the salary, in order to get better chances of approval. In this ongoing work, we propose to discuss, with a first proposition, the issue of detecting a potential local adversarial example on classical tabular data by providing to a human expert the locally critical features for the classifier's decision, in order to control the provided information and avoid a fraud.
38.
Ruf, Boris; Sammarco, Matteo; Detyniecki, Marcin
Contract Statements Knowledge Service for Chatbots Journal Article
In: IEEE Systems, Man, and Cybernetics, 2019.
BibTeX | Tags: Human-Computer Interface & HCXAI
@article{ruf2019,
title = {Contract Statements Knowledge Service for Chatbots},
author = {Boris Ruf and Matteo Sammarco and Marcin Detyniecki},
year = {2019},
date = {2019-01-01},
journal = {IEEE Systems, Man, and Cybernetics},
keywords = {Human-Computer Interface & HCXAI},
pubstate = {published},
tppubtype = {article}
}
39.
Laugel, Thibault; Renard, Xavier; Lesot, Marie-Jeanne; Marsala, Christophe; Detyniecki, Marcin
Defining Locality for Surrogates in Post-hoc Interpretablity Proceedings Article
In: Workshop on Human Interpretability for Machine Learning (WHI)-International Conference on Machine Learning (ICML), 2018.
Abstract | Links | BibTeX | Tags: Explainability & Interpretability
@inproceedings{laugel2018defining,
title = {Defining Locality for Surrogates in Post-hoc Interpretablity},
author = {Thibault Laugel and Xavier Renard and Marie-Jeanne Lesot and Christophe Marsala and Marcin Detyniecki},
url = {https://arxiv.org/abs/1806.07498},
year = {2018},
date = {2018-01-01},
booktitle = {Workshop on Human Interpretability for Machine Learning (WHI)-International Conference on Machine Learning (ICML)},
abstract = {Local surrogate models, to approximate the local decision boundary of a black-box classifier, constitute one approach to generate explanations for the rationale behind an individual prediction made by the back-box. This paper highlights the importance of defining the right locality, the neighborhood on which a local surrogate is trained, in order to approximate accurately the local black-box decision boundary. Unfortunately, as shown in this paper, this issue is not only a parameter or sampling distribution challenge and has a major impact on the relevance and quality of the approximation of the local black-box decision boundary and thus on the meaning and accuracy of the generated explanation. To overcome the identified problems, quantified with an adapted measure and procedure, we propose to generate surrogate-based explanations for individual predictions based on a sampling centered on particular place of the decision boundary, relevant for the prediction to be explained, rather than on the prediction itself as it is classically done. We evaluate the novel approach compared to state-of-the-art methods and a straightforward improvement thereof on four UCI datasets.},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {inproceedings}
}
Local surrogate models, to approximate the local decision boundary of a black-box classifier, constitute one approach to generate explanations for the rationale behind an individual prediction made by the back-box. This paper highlights the importance of defining the right locality, the neighborhood on which a local surrogate is trained, in order to approximate accurately the local black-box decision boundary. Unfortunately, as shown in this paper, this issue is not only a parameter or sampling distribution challenge and has a major impact on the relevance and quality of the approximation of the local black-box decision boundary and thus on the meaning and accuracy of the generated explanation. To overcome the identified problems, quantified with an adapted measure and procedure, we propose to generate surrogate-based explanations for individual predictions based on a sampling centered on particular place of the decision boundary, relevant for the prediction to be explained, rather than on the prediction itself as it is classically done. We evaluate the novel approach compared to state-of-the-art methods and a straightforward improvement thereof on four UCI datasets.
40.
Laugel, Thibault; Lesot, Marie-Jeanne; Marsala, Christophe; Renard, Xavier; Detyniecki, Marcin
Comparison-based Inverse Classification for Interpretability in Machine Learning Proceedings Article
In: 17th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU 2018), pp. 100–111, Springer Verlag 2018.
BibTeX | Tags: Explainability & Interpretability
@inproceedings{laugel2018comparison,
title = {Comparison-based Inverse Classification for Interpretability in Machine Learning},
author = {Thibault Laugel and Marie-Jeanne Lesot and Christophe Marsala and Xavier Renard and Marcin Detyniecki},
year = {2018},
date = {2018-01-01},
booktitle = {17th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU 2018)},
pages = {100--111},
organization = {Springer Verlag},
keywords = {Explainability & Interpretability},
pubstate = {published},
tppubtype = {inproceedings}
}