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

Although neural module networks have an architectural bias towardscompositionality, they require gold standard layouts to generalizesystematically in practice. When instead learning layouts and modules jointly,compositionality does not arise automatically and an explicit pressure isnecessary for the emergence of layouts exhibiting the right structure. Wepropose to address this problem using iterated learning, a cognitive sciencetheory of the emergence of compositional languages in nature that has primarilybeen applied to simple referential games in machine learning. Considering thelayouts of module networks as samples from an emergent language, we useiterated learning to encourage the development of structure within thislanguage. We show that the resulting layouts support systematic generalizationin neural agents solving the more complex task of visual question-answering.Our regularized iterated learning method can outperform baselines withoutiterated learning on SHAPES-SyGeT (SHAPES Systematic Generalization Test), anew split of the SHAPES dataset we introduce to evaluate systematicgeneralization, and on CLOSURE, an extension of CLEVR also designed to testsystematic generalization. We demonstrate superior performance in recoveringground-truth compositional program structure with limited supervision on bothSHAPES-SyGeT and CLEVR.

Iterated learning for emergent systematicity in VQA