Tweeted By @rasbt
audioLIME: Local Interpretable Model-agnostic Explanations for waveforms & spectrograms. @veroamilbe
— Sebastian Raschka (@rasbt) August 8, 2020
et al decompose input audio sources into multiple components (both sources and segments) to produce explanations that are listenable: https://t.co/UhdklyYSwt pic.twitter.com/kTdlFgICb1