Interpretable Machine Learning (IML) / Explainable AI (XAI)
Machine learning models are often referred to as black boxes because their predictions are often intransparent and not easy to understand for humans. Numerous post-hoc methods from the field of interpretable machine learning have been developed in recent years to gain new insights into black box models and their underlying data. Furthermore, model interpretation helps to validate and debug models, which also contributes to a better understanding. In our research group, we explore and implement approaches for interpretable machine learning. Our research focus lies on model-agnostic methods for tabular data.
Focus Areas:
- Analysis of limitations of interpretation methods
- Connection between causality and model interpretation
- Connection between sensitivity analysis and model interpretation
- Counterfactual explanations
- Consolidation of state-of-the-art methods
- Interpretable boosting models
- Partial dependence plots and accumulated local effects
- Permutation feature importance
- Shapley values
Projects and Software
-
iml: R package for model-agnostic interpretability methods. -
Book: Interpretable Machine Learning: An open source book that may serve as a guide for making black box models explainable.
Members
| Name | Position | |||
|---|---|---|---|---|
| Dr. Giuseppe Casalicchio | PostDoc / Lead | |||
| Gunnar König | PhD Student | |||
| Christian Scholbeck | PhD Student | |||
| Susanne Dandl | PhD Student | |||
| Julia Herbinger | PhD Student | |||
| Fiona Katharina Ewald | PhD Student |
Former Members
| Name | Year | |||
|---|---|---|---|---|
| Christoph Molnar | 2017 – 2021 | |||
| Quay Au | 2017 – 2020 |
Publications
- Herbinger J, Bischl B, Casalicchio G (2022) REPID: Regional Effect Plots with implicit Interaction Detection. International Conference on Artificial Intelligence and Statistics (AISTATS) 25.
link | pdf. - Scholbeck CA, Casalicchio G, Molnar C, Bischl B, Heumann C (2022) Marginal Effects for Non-Linear Prediction Functions.
link . - König G, Freiesleben T, Bischl B, Casalicchio G, Grosse-Wentrup M (2021) Decomposition of Global Feature Importance into Direct and Associative Components (DEDACT).
link. - König G, Molnar C, Bischl B, Grosse-Wentrup M (2021) Relative Feature Importance 2020 25th International Conference on Pattern Recognition (ICPR), pp. 9318–9325.
link | pdf . - Molnar C, Freiesleben T, König G, Casalicchio G, Wright MN, Bischl B (2021) Relating the Partial Dependence Plot and Permutation Feature Importance to the Data Generating Process. arXiv preprint arXiv:2109.01433.
link | pdf. - Moosbauer J, Herbinger J, Casalicchio G, Lindauer M, Bischl B (2021) Explaining Hyperparameter Optimization via Partial Dependence Plots. Advances in Neural Information Processing Systems (NeurIPS 2021) 34.
link | pdf. - Moosbauer J, Herbinger J, Casalicchio G, Lindauer M, Bischl B (2021) Towards Explaining Hyperparameter Optimization via Partial Dependence Plots 8th ICML Workshop on Automated Machine Learning (AutoML),
link | pdf. - Au Q, Herbinger J, Stachl C, Bischl B, Casalicchio G (2021) Grouped Feature Importance and Combined Features Effect Plot. arXiv preprint arXiv:2104.11688.
link | pdf. - Dandl S, Molnar C, Binder M, Bischl B (2020) Multi-Objective Counterfactual Explanations. In: In: Bäck T , In: Preuss M , In: Deutz A et al. (eds) Parallel Problem Solving from Nature – PPSN XVI, pp. 448–469. Springer International Publishing, Cham.
link. - Scholbeck CA, Molnar C, Heumann C, Bischl B, Casalicchio G (2020) Sampling, Intervention, Prediction, Aggregation: A Generalized Framework for Model-Agnostic Interpretations. In: In: Cellier P , In: Driessens K (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2019, pp. 205–216. Springer International Publishing, Cham.
link | pdf. - Molnar C, König G, Bischl B, Casalicchio G (2020) Model-agnostic Feature Importance and Effects with Dependent Features–A Conditional Subgroup Approach. arXiv preprint arXiv:2006.04628.
link | pdf. - Molnar C, König G, Herbinger J et al. (2020) Pitfalls to Avoid when Interpreting Machine Learning Models. ICML workshop XXAI: Extending Explainable AI Beyond Deep Models and Classifiers.
link | pdf. - Molnar C, Casalicchio G, Bischl B (2020) Quantifying Model Complexity via Functional Decomposition for Better Post-hoc Interpretability. In: In: Cellier P , In: Driessens K (eds) Machine Learning and Knowledge Discovery in Databases, pp. 193–204. Springer International Publishing, Cham. link | pdf.
- Molnar C, Casalicchio G, Bischl B (2020) Interpretable Machine Learning – A Brief History, State-of-the-Art and Challenges. In: In: Koprinska I , In: Kamp M , In: Appice A et al. (eds) ECML PKDD 2020 Workshops, pp. 417–431. Springer International Publishing, Cham.
link | pdf . - Liew B, Rügamer D, De Nunzio A, Falla D (2020) Interpretable machine learning models for classifying low back pain status using functional physiological variables. European Spine Journal 29, 1845–1859.
link. - Casalicchio G, Molnar C, Bischl B (2019) Visualizing the Feature Importance for Black Box Models. In: In: Berlingerio M , In: Bonchi F , In: Gärtner T , In: Hurley N , In: Ifrim G (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2018, pp. 655–670. Springer International Publishing, Cham.
link | pdf. - Molnar C, Casalicchio G, Bischl B (2018) iml: An R package for Interpretable Machine Learning. The Journal of Open Source Software 3, 786. link | pdf.
- Casalicchio G, Bischl B, Boulesteix A-L, Schmid M (2015) The residual-based predictiveness curve: A visual tool to assess the performance of prediction models. Biometrics 72, 392–401.
link | pdf.
Contact
Feel free to contact us if you are looking for collaborations:
giuseppe.casalicchio [at] stat.uni-muenchen.de