Understanding Inhibition through Maximally Tense Images

Christopher J Hamblin, Srijani Saha, Talia Konkle, George A. Alvarez

Keywords: Object recognition, Feature Inhibition, Feature Visualization, Superposition

TL;DR: We study the functional role of negative weights in object recognition models by looking for images that yield excitatory and inhibitory attributions simultaneously

Abstract: 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 induce the same attribution motif as maximally tense images.

Submission Number: 88