Cognitome: Neural Hypernetworks and Percolation Hypothesis of Consciousness Konstantin Anokhin (Moscow State University, Cente, NRC Kurchatov Institute, Moscow, Russia, Moscow, Russian Federation) C12
Despite impressive advances in neuroscience, the nature of consciousness and other higher brain functions still eludes satisfactory understanding. This is not due to lack of experimental facts - thousands of papers are published annually in cognitive neuroscience. What is missing is not just more facts, but rather a theoretical framework that could make sense of them and relate mental phenomena to brain functions. This situation, known as explanatory gap, calls for new explanatory models and principles. The present report proposes such new principle of organization of higher brain functions - the neural hypernetwork principle. It suggests that evolution, development and learning shape neural networks (connectome) into a higher-order structures - cognitive neural hypernetworks (cognitome). While traditional neural network approach focuses on pairwise relations between elements, hypernetwork is a concept from the algebraic topology that describes relations between many elements. Hypernetworks are formed as networks from relational simplices, otherwise called hypersimplices, in which the relational structure between the lower-level elements is explicit. Hypernetwork theory thus provides formalism for the multilevel systems, enables formulation of theoretical relationships between micro and macro levels, gives description of emergent phenomena in the multilevel systems, and allows modeling much more complex structures than networks and hypergraphs. Hypernetwork brain theory (HBT) argues that each brain at its maximal causal power is cognitome - a neural hypernetwork with emergent cognitive properties. Vertices of cognitome called COGs (GOgnitive Groups) are subsets of elements from the underlying neuronal network that are associated by a common experience. Being stored in long-term memory they represent units of knowledge of a cognitive agent. In terms of algebraic topology COG is hypersimplex. Its base is formed from the vertices of underlying neural network and its apex is a vertex in a hypernetwork that possess an emergent cognitive quality. Edges between COGs are called LOCs (Links Of Cogs), are encoded by overlapping neurons of linked COGs and represent units of causal knowledge of a cognitive agent. HBT proposes three foundational principles for the origin of cognitive neural hypernetwork, three rules of there generation, and describes various mental processes as different forms of traffic in this hypernetwork. Consciousness is suggested to be a particular form of this traffic characterized by hypernetwork percolation that results in global access to COGs of the cognitome. HBT makes specific predictions to test these propositions.