Suggested critical behaviour of a biologically inspired neural network model
Joseph Rushton Wakeling · arXiv (Cornell University) · 2002
It has been suggested by Bak, Stassinopoulos, Chialvo and others that the brain operates at a critical level. In this letter, we examine a neural network proposed to model this, the simple three-layer version of the Chialvo-Bak ‘minibrain ’ [Neuroscience 90, 1137 (1999)], and present numerical results which suggest that it does indeed display critical (scale-free) behaviour (1/f β noise, with β ≈ 1.3) subject to the tuning of a control parameter, ηC = ηIM/(ηIP(ηOP −1)), where ηIP, ηIM and ηOP are, respectively, the sizes of the input, intermediary and output layers of neurons. The cause of this behaviour is identified as interference between active paths in the network, and an heuristic analogy is drawn with Bak, Chen and Tang’s forest-fire model.