This paper discusses the process of concept learning in humans, a crucial part of cognition that affects categorization, reasoning, memory, and decision making. It reviews findings from computational neuroscience and cognitive psychology, highlighting two types of brain concept representations: multisensory and text-derived. These representations are coordinated by a semantic control system. The paper presents a computational model for concept learning inspired by these mechanisms, using spiking neural networks. This model effectively overcomes the challenges of diverse source inputs and imbalanced dimensionality, resulting in human-like concept representations.
Publication date: 15 Jan 2024
Project Page: https://arxiv.org/abs/2401.06471v1
Paper: https://arxiv.org/pdf/2401.06471
