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

An important challenge in neuroevolution is to evolve complex neural networkswith multiple modes of behavior. Indirect encodings can potentially answer thischallenge. Yet in practice, indirect encodings do not yield effectivemultimodal controllers. Thus, this paper introduces novel multimodal extensionsto HyperNEAT, a popular indirect encoding. A previous multimodal HyperNEATapproach called situational policy geometry assumes that multiple brainsbenefit from being embedded within an explicit geometric space. However,experiments here illustrate that this assumption unnecessarily constrainsevolution, resulting in lower performance. Specifically, this paper introducesHyperNEAT extensions for evolving many brains without assuming geometricrelationships between them. The resulting Multi-Brain HyperNEAT can exploithuman-specified task divisions to decide when each brain controls the agent, orcan automatically discover when brains should be used, by means of preferenceneurons. A further extension called module mutation allows evolution todiscover the number of brains, enabling multimodal behavior with even lessexpert knowledge. Experiments in several multimodal domains highlight thatmulti-brain approaches are more effective than HyperNEAT without multimodalextensions, and show that brains without a geometric relation to each otheroutperform situational policy geometry. The conclusion is that Multi-BrainHyperNEAT provides several promising techniques for evolving complex multimodalbehavior.

Using Indirect Encoding of Multiple Brains to Produce Multimodal Behavior