English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Despite the remarkable success of deep neural networks in a myriad ofsettings, several works have demonstrated their overwhelming sensitivity tonear-imperceptible perturbations, known as adversarial attacks. On the otherhand, prior works have also observed that deep networks can be under-sensitive,wherein large-magnitude perturbations in input space do not induce appreciablechanges to network activations. In this work, we study in detail the phenomenonof under-sensitivity in vision models such as CNNs and Transformers, andpresent techniques to study the geometry and extent of "equi-confidence" levelsets of such networks. We propose a Level Set Traversal algorithm thatiteratively explores regions of high confidence with respect to the input spaceusing orthogonal components of the local gradients. Given a source image, weuse this algorithm to identify inputs that lie in the same equi-confidencelevel set as the source image despite being perceptually similar to arbitraryimages from other classes. We further observe that the source image is linearlyconnected by a high-confidence path to these inputs, uncovering a star-likestructure for level sets of deep networks. Furthermore, we attempt to identifyand estimate the extent of these connected higher-dimensional regions overwhich the model maintains a high degree of confidence. The code for thisproject is publicly available athttps://github.com/SriramB-98/blindspots-neurips-sub

Exploring Geometry of Blind Spots in Vision Models