Skip to Main content Skip to Navigation
Preprints, Working Papers, ...

Don't Lie to Me! Robust and Efficient Explainability with Verified Perturbation Analysis

Abstract : A variety of methods have been proposed to try to explain how deep neural networks make their decisions. Key to those approaches is the need to sample the pixel space efficiently in order to derive importance maps. However, it has been shown that the sampling methods used to date introduce biases and other artifacts, leading to inaccurate estimates of the importance of individual pixels and severely limit the reliability of current explainability methods. Unfortunately, the alternative-to exhaustively sample the image space is computationally prohibitive. In this paper, we introduce EVA (Explaining using Verified perturbation Analysis)-the first explainability method guarantee to have an exhaustive exploration of a perturbation space. Specifically, we leverage the beneficial properties of verified perturbation analysis-time efficiency, tractability and guaranteed complete coverage of a manifold-to efficiently characterize the input variables that are most likely to drive the model decision. We evaluate the approach systematically and demonstrate state-of-the-art results on multiple benchmarks.
Document type :
Preprints, Working Papers, ...
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-03576300
Contributor : Thomas FEL Connect in order to contact the contributor
Submitted on : Tuesday, February 15, 2022 - 10:55:08 PM
Last modification on : Monday, April 4, 2022 - 4:16:04 PM
Long-term archiving on: : Monday, May 16, 2022 - 9:01:28 PM

File

Formal_and_Explainability.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03576300, version 1

Collections

Citation

Thomas Fel, Mélanie Ducoffe, David Vigouroux, Rémi Cadène, Mikael Capelle, et al.. Don't Lie to Me! Robust and Efficient Explainability with Verified Perturbation Analysis. 2022. ⟨hal-03576300⟩

Share

Metrics

Record views

7

Files downloads

3