A toolbox to iNNvestigate neural networks' predictions!
(a.k.a. Path-Integrated Gradients, a.k.a. Axiomatic Attribution for Deep Networks)
DeepLIFT: Deep Learning Important FeaTures
Die Hauptfunktionen von kundajelab/deeplift sind: Explainable AI Libraries, Model Interpretability.
Open-Source-Alternativen zu kundajelab/deeplift sind unter anderem: austinrochford/pycebox — ⬛ Python Individual Conditional Expectation Plot Toolbox. aerdem4/lofo-importance — Leave One Feature Out Importance. albermax/innvestigate — A toolbox to iNNvestigate neural networks' predictions! andosa/treeinterpreter — TreeInterpreter. ankurtaly/integrated-gradients — (a.k.a. Path-Integrated Gradients, a.k.a. Axiomatic Attribution for Deep Networks). csinva/imodels — Interpretable ML package 🔍 for concise, transparent, and accurate predictive modeling (sklearn-compatible).
