On Baselines for Local Feature Attributions

Johannes Haug, Stefan Zürn, Peter El-Jiz +1

High-performing predictive models, such as neural nets, usually operate as black boxes, which raises serious concerns about their interpretability. Local feature attribution method…