Explainable empirical risk minimization

Loading...
Thumbnail Image

Access rights

openAccess
publishedVersion

URL

Journal Title

Journal ISSN

Volume Title

A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä

Date

2024-03

Major/Subject

Mcode

Degree programme

Language

en

Pages

Series

Neural Computing and Applications, Volume 36, issue 8, pp. 3983-3996

Abstract

The successful application of machine learning (ML) methods increasingly depends on their interpretability or explainability. Designing explainable ML (XML) systems is instrumental for ensuring transparency of automated decision-making that targets humans. The explainability of ML methods is also an essential ingredient for trustworthy artificial intelligence. A key challenge in ensuring explainability is its dependence on the specific human end user of an ML system. The users of ML methods might have vastly different background knowledge about ML principles, with some having formal training in the specific field and others having none. We use information-theoretic concepts to develop a novel measure for the subjective explainability of predictions delivered by a ML method. We construct this measure via the conditional entropy of predictions, given the user signal. Our approach allows for a wide range of user signals, ranging from responses to surveys to biophysical measurements. We use this measure of subjective explainability as a regularizer for model training. The resulting explainable empirical risk minimization (EERM) principle strives to balance subjective explainability and risk. The EERM principle is flexible and can be combined with arbitrary ML models. We present several practical implementations of EERM for linear models and decision trees. Numerical experiments demonstrate the application of EERM to weather prediction and detecting inappropriate language in social media.

Description

Funding Information: Open Access funding provided by Aalto University. The work in this paper is supported by the following three fundings: funding “Ensemble nowcasting of irradiance/clouds for solar energy using novel machine learning tools—can AI beat physics?" with project number 885377 granted by Austrian Research Promotion Agency (FFG) in 2021, funding “Intelligent Techniques in Condition Monitoring of Electromechanical Energy Conversion Systems" granted by Academy of Finland (decision number 331197) in 2020, and funding“XAI-based software-defined energy networks via packetized management for fossil fuel-free next-generation of industrial cyber-physical systems (X-SDEN)” granted by Academy of Finland (decision number 349966) in 2022. Publisher Copyright: © 2023, The Author(s).

Keywords

Empirical risk minimization, Explainable machine learning, Subjective explainability

Other note

Citation

Zhang, L, Karakasidis, G, Odnoblyudova, A, Dogruel, L, Tian, Y & Jung, A 2024, ' Explainable empirical risk minimization ', Neural Computing and Applications, vol. 36, no. 8, pp. 3983-3996 . https://doi.org/10.1007/s00521-023-09269-3