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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Despite their impressive performance in object recognition and other tasksunder standard testing conditions, deep networks often fail to generalize toout-of-distribution (o.o.d.) samples. One cause for this shortcoming is thatmodern architectures tend to rely on "shortcuts" - superficial features thatcorrelate with categories without capturing deeper invariants that hold acrosscontexts. Real-world concepts often possess a complex structure that can varysuperficially across contexts, which can make the most intuitive and promisingsolutions in one context not generalize to others. One potential way to improveo.o.d. generalization is to assume simple solutions are unlikely to be validacross contexts and avoid them, which we refer to as the too-good-to-be-trueprior. A low-capacity network (LCN) with a shallow architecture should only beable to learn surface relationships, including shortcuts. We find that LCNs canserve as shortcut detectors. Furthermore, an LCN's predictions can be used in atwo-stage approach to encourage a high-capacity network (HCN) to rely on deeperinvariant features that should generalize broadly. In particular, items thatthe LCN can master are downweighted when training the HCN. Using a modifiedversion of the CIFAR-10 dataset in which we introduced shortcuts, we found thatthe two-stage LCN-HCN approach reduced reliance on shortcuts and facilitatedo.o.d. generalization.

A Too-Good-to-be-True Prior to Reduce Shortcut Reliance