Computational processes that appear to model human memory (Q2364899)

In this paper the author promotes the idea of the usage of two different types of networks for the modelling of neural behaviour and neural states. For the former a directed, asymmetric model is proposed, and for the latter a bi-directional, symmetric interconnection between the nodes/neurones is suggested; the author focuses on the symmetric relationships modelled as graphs or networks and reduces them to a unique, irreducible trace using closures, similar to long term memory consolidation. First, the mathematical concepts of sets, relations and closures are introduced, followed by a description of irreducible networks. Next, the author presents an approach for evaluating the similarity of networks, from a structural angle. The paper concludes with a translation of these concepts to biological memory focussing on the role of closures in a biological context and on the role of chordless cycles. For the entire collection see [Zbl 1365.92002].