Abstract
Artificial neural networks (ANNs) were inspired by the architecture andfunctions of the human brain and have revolutionised the field of artificialintelligence (AI). Inspired by studies on the latent geometry of the brain, inthis perspective paper we posit that an increase in the research andapplication of hyperbolic geometry in ANNs and machine learning will lead toincreased accuracy, improved feature space representations and more efficientmodels across a range of tasks. We examine the structure and functions of thehuman brain, emphasising the correspondence between its scale-free hierarchicalorganization and hyperbolic geometry, and reflecting on the central rolehyperbolic geometry plays in facilitating human intelligence. Empiricalevidence indicates that hyperbolic neural networks outperform Euclidean modelsfor tasks including natural language processing, computer vision and complexnetwork analysis, requiring fewer parameters and exhibiting bettergeneralisation. Despite its nascent adoption, hyperbolic geometry holds promisefor improving machine learning models through brain-inspired geometricrepresentations.