Hippocluster: An efficient, hippocampus-inspired algorithm for graph clustering

Random walks can reveal communities/clusters in networks, because they are more likely to stay within a cluster than leave it. Thus, one family of community detection algorithms uses random measure distance between nodes, and then applies clustering methods these distances. Interestingly, information processing the brain may suggest simpler method learning clusters directly from walks. Drawing inspiration hippocampus, structure involved navigation, we propose two-layer neural framework. Neurons layer associated with graph nodes activated by These activations cause neurons second become tuned through simple associative learning. The system be modelled as Online Spherical K-Means applied novel walk-space where all equidistant each other. In tests on benchmark real-world graphs, our framework achieved high normalized mutual scores, executed faster comparator algorithms, showed data efficiency requiring few 6 per node. Biological systems known for adaptability; here, drawing has provided properties.