Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fel, Thomas, Boissin, Thibaut, Boutin, Victor, Picard, Agustin, Novello, Paul, Colin, Julien, Linsley, Drew, Rousseau, Tom, Cadène, Rémi, Goetschalckx, Lore, Gardes, Laurent, Serre, Thomas
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909422554447872
author Fel, Thomas
Boissin, Thibaut
Boutin, Victor
Picard, Agustin
Novello, Paul
Colin, Julien
Linsley, Drew
Rousseau, Tom
Cadène, Rémi
Goetschalckx, Lore
Gardes, Laurent
Serre, Thomas
author_facet Fel, Thomas
Boissin, Thibaut
Boutin, Victor
Picard, Agustin
Novello, Paul
Colin, Julien
Linsley, Drew
Rousseau, Tom
Cadène, Rémi
Goetschalckx, Lore
Gardes, Laurent
Serre, Thomas
contents Feature visualization has gained substantial popularity, particularly after the influential work by Olah et al. in 2017, which established it as a crucial tool for explainability. However, its widespread adoption has been limited due to a reliance on tricks to generate interpretable images, and corresponding challenges in scaling it to deeper neural networks. Here, we describe MACO, a simple approach to address these shortcomings. The main idea is to generate images by optimizing the phase spectrum while keeping the magnitude constant to ensure that generated explanations lie in the space of natural images. Our approach yields significantly better results (both qualitatively and quantitatively) and unlocks efficient and interpretable feature visualizations for large state-of-the-art neural networks. We also show that our approach exhibits an attribution mechanism allowing us to augment feature visualizations with spatial importance. We validate our method on a novel benchmark for comparing feature visualization methods, and release its visualizations for all classes of the ImageNet dataset on https://serre-lab.github.io/Lens/. Overall, our approach unlocks, for the first time, feature visualizations for large, state-of-the-art deep neural networks without resorting to any parametric prior image model.
format Preprint
id arxiv_https___arxiv_org_abs_2306_06805
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization
Fel, Thomas
Boissin, Thibaut
Boutin, Victor
Picard, Agustin
Novello, Paul
Colin, Julien
Linsley, Drew
Rousseau, Tom
Cadène, Rémi
Goetschalckx, Lore
Gardes, Laurent
Serre, Thomas
Computer Vision and Pattern Recognition
Artificial Intelligence
Feature visualization has gained substantial popularity, particularly after the influential work by Olah et al. in 2017, which established it as a crucial tool for explainability. However, its widespread adoption has been limited due to a reliance on tricks to generate interpretable images, and corresponding challenges in scaling it to deeper neural networks. Here, we describe MACO, a simple approach to address these shortcomings. The main idea is to generate images by optimizing the phase spectrum while keeping the magnitude constant to ensure that generated explanations lie in the space of natural images. Our approach yields significantly better results (both qualitatively and quantitatively) and unlocks efficient and interpretable feature visualizations for large state-of-the-art neural networks. We also show that our approach exhibits an attribution mechanism allowing us to augment feature visualizations with spatial importance. We validate our method on a novel benchmark for comparing feature visualization methods, and release its visualizations for all classes of the ImageNet dataset on https://serre-lab.github.io/Lens/. Overall, our approach unlocks, for the first time, feature visualizations for large, state-of-the-art deep neural networks without resorting to any parametric prior image model.
title Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2306.06805