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

Vision transformers (ViTs) are top performing models on many computer visionbenchmarks and can accurately predict human behavior on object recognitiontasks. However, researchers question the value of using ViTs as models ofbiological learning because ViTs are thought to be more data hungry thanbrains, with ViTs requiring more training data to reach similar levels ofperformance. To test this assumption, we directly compared the learningabilities of ViTs and animals, by performing parallel controlled rearingexperiments on ViTs and newborn chicks. We first raised chicks in impoverishedvisual environments containing a single object, then simulated the trainingdata available in those environments by building virtual animal chambers in avideo game engine. We recorded the first-person images acquired by agentsmoving through the virtual chambers and used those images to train selfsupervised ViTs that leverage time as a teaching signal, akin to biologicalvisual systems. When ViTs were trained through the eyes of newborn chicks, theViTs solved the same view invariant object recognition tasks as the chicks.Thus, ViTs were not more data hungry than newborn visual systems: both learnedview invariant object representations in impoverished visual environments. Theflexible and generic attention based learning mechanism in ViTs combined withthe embodied data streams available to newborn animals appears sufficient todrive the development of animal-like object recognition.

Are Vision Transformers More Data Hungry Than Newborn Visual Systems?