Topic: relational learning

Traditional brain-style (connectionist) approaches basically hit a wall when it comes to relational cognition. As an alternative to the well-known approaches of structured connectionism and deep learning, I present an engine for relational pattern recognition based on minimalist reinterpretations of first principles of connectionism. Results of computational experiments will be discussed on problems testing relational learning and universal generalization.

Humans can infer never-experienced global relational structures (e.g., "A>D”) by observing and integrating multiple local relations (e.g., "A>B", "B>C", “C>D”). This capability for “transitive inference” (TI) depends on the medial temporal lobe (MTL). But the mechanisms by which the MTL performs TI remain poorly understood. Previous work suggested that “concept cells” (CCs), which are neurons that selectively respond to preferred concepts (e.g. “A”), might contribute to relational learning and potentially TI by widening their tuning to new concepts that co-occurred with initially r