A Rate-Distortion Framework for Explaining Black-Box Model Decisions

Stefan Kolek, Duc Anh Nguyen, Ron Levie, Joan Bruna, Gitta Kutyniok

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

Abstract

We present the Rate-Distortion Explanation (RDE) framework, a mathematically well-founded method for explaining black-box model decisions. The framework is based on perturbations of the target input signal and applies to any differentiable pre-trained model such as neural networks. Our experiments demonstrate the framework’s adaptability to diverse data modalities, particularly images, audio, and physical simulations of urban environments.

Original languageEnglish
Title of host publicationxxAI@ICML
EditorsAndreas Holzinger, Randy Goebel, Ruth Fong, Taesup Moon, Klaus-Robert Müller, Wojciech Samek
PublisherSpringer Science and Business Media Deutschland GmbH
Pages91-115
Number of pages25
ISBN (Print)9783031040825
DOIs
StatePublished - 2022
EventInternational Workshop on Extending Explainable AI Beyond Deep Models and Classifiers, xxAI 2020, held in Conjunction with ICML 2020 - Vienna, Austria
Duration: 18 Jul 202018 Jul 2020

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13200 LNAI

Conference

ConferenceInternational Workshop on Extending Explainable AI Beyond Deep Models and Classifiers, xxAI 2020, held in Conjunction with ICML 2020
Country/TerritoryAustria
CityVienna
Period18/07/2018/07/20

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'A Rate-Distortion Framework for Explaining Black-Box Model Decisions'. Together they form a unique fingerprint.

Cite this