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

Striking progress has recently been made in understanding human cognition byanalyzing how its neuronal underpinnings are engaged in different modes ofinformation processing. Specifically, neural information can be decomposed intosynergistic, redundant, and unique features, with synergistic components beingparticularly aligned with complex cognition. However, two fundamental questionsremain unanswered: (a) precisely how and why a cognitive system can becomehighly synergistic; and (b) how these informational states map onto artificialneural networks in various learning modes. To address these questions, here weemploy an information-decomposition framework to investigate the informationprocessing strategies adopted by simple artificial neural networks performing avariety of cognitive tasks in both supervised and reinforcement learningsettings. Our results show that synergy increases as neural networks learnmultiple diverse tasks. Furthermore, performance in tasks requiring integrationof multiple information sources critically relies on synergistic neurons.Finally, randomly turning off neurons during training through dropout increasesnetwork redundancy, corresponding to an increase in robustness. Overall, ourresults suggest that while redundant information is required for robustness toperturbations in the learning process, synergistic information is used tocombine information from multiple modalities -- and more generally for flexibleand efficient learning. These findings open the door to new ways ofinvestigating how and why learning systems employ specificinformation-processing strategies, and support the principle that the capacityfor general-purpose learning critically relies in the system's informationdynamics.

Synergistic information supports modality integration and flexible learning in neural networks solving multiple tasks