Interpretable generalized additive neural networks

Mathias Kraus, Daniel Tschernutter, Sven Weinzierl, Patrick Zschech

Abstract: Highlights•We introduce a new ML model IGANN for high performance & interpretability.•We propose an efficient training algorithm with extreme learning machines.•Our training algorithm scales linearly with the training dataset.•We demonstrate favorable results in a variety of numerical experiments.•We demonstrate application on three real-world case studies.