Saved in:
Bibliographic Details
Main Authors: Hamblin, Chris, Saha, Srijani, Konkle, Talia, Alvarez, George
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.05598
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914830052491264
author Hamblin, Chris
Saha, Srijani
Konkle, Talia
Alvarez, George
author_facet Hamblin, Chris
Saha, Srijani
Konkle, Talia
Alvarez, George
contents We address the functional role of 'feature inhibition' in vision models; that is, what are the mechanisms by which a neural network ensures images do not express a given feature? We observe that standard interpretability tools in the literature are not immediately suited to the inhibitory case, given the asymmetry introduced by the ReLU activation function. Given this, we propose inhibition be understood through a study of 'maximally tense images' (MTIs), i.e. those images that excite and inhibit a given feature simultaneously. We show how MTIs can be studied with two novel visualization techniques; +/- attribution inversions, which split single images into excitatory and inhibitory components, and the attribution atlas, which provides a global visualization of the various ways images can excite/inhibit a feature. Finally, we explore the difficulties introduced by superposition, as such interfering features induce the same attribution motif as MTIs.
format Preprint
id arxiv_https___arxiv_org_abs_2406_05598
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Understanding Inhibition Through Maximally Tense Images
Hamblin, Chris
Saha, Srijani
Konkle, Talia
Alvarez, George
Computer Vision and Pattern Recognition
We address the functional role of 'feature inhibition' in vision models; that is, what are the mechanisms by which a neural network ensures images do not express a given feature? We observe that standard interpretability tools in the literature are not immediately suited to the inhibitory case, given the asymmetry introduced by the ReLU activation function. Given this, we propose inhibition be understood through a study of 'maximally tense images' (MTIs), i.e. those images that excite and inhibit a given feature simultaneously. We show how MTIs can be studied with two novel visualization techniques; +/- attribution inversions, which split single images into excitatory and inhibitory components, and the attribution atlas, which provides a global visualization of the various ways images can excite/inhibit a feature. Finally, we explore the difficulties introduced by superposition, as such interfering features induce the same attribution motif as MTIs.
title Understanding Inhibition Through Maximally Tense Images
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2406.05598