Interpreting Classifiers through Attribute Interactions in Datasets
By Andreas Henelius, Kai Puolamäki, Antti Ukkonen. 2017
40 citations recorded by the source corpus.
In this work we present the novel ASTRID method for investigating which attribute interactions classifiers exploit when making predictions. Attribute interactions in classification tasks mean that two or more attributes together provide stronger evidence for a particular class label. Knowledge of such interactions makes models more interpretable by revealing associations between attributes. This has applications, e.g., in pharmacovigilance to identify interactions between drugs or in bioinformatics to investigate associations between single nucleotide polymorphisms. We also show how the found attribute partitioning is related to a factorisation of the data generating distribution and empirically demonstrate the utility of the proposed method.
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Francisco Herrera
- Salvador Garcia
- Javier Del Ser
- Daniel Molina
- Alejandro Barredo Arrieta
- Adrien Bennetot
- Natalia Díaz Rodríguez
- Nadia Burkart
Citations in the corpus
- Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, positive): “The study in [123] presents the so-called Automatic STRucture IDentification method (ASTRID) to inspect which attributes are exploited by a classifier to generate a prediction.”
- Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI) (2018, methodology): “The academic AI research community is well represented in this application domain with the DAPRA Ambitious XAI program, alongwith some research initiatives that studies explainability in this domain [54].”
- Principles and Practice of Explainable Machine Learning (2020, methodology): “Likewise, the work in [36] presents a method (ASTRID) that aims at identifying which attributes are utilized by a classifier in prediction time.”
- MAGIX: Model Agnostic Globally Interpretable Explanations (2017, unknown): “, 2017), (Kim & Seo, 2017), (Koh & Liang, 2017), (Henelius et al., 2017), (Phillips et al.”
- A Survey on the Explainability of Supervised Machine Learning (2020, background): “Features in the same group interact; features in different groups do not interact (Henelius et al., 2017).”