作者: Shun-ichi Amari
DOI: 10.1007/978-3-642-54593-1_3
关键词:
摘要: Mathematical treatments of the dynamics neural fields become much simpler when Heaviside function is used as an activation function. This because excited or active region reduce to boundary. We call this regime world. Here, we visit world and briefly review bump in 1D, 1D two-layer, 2D cases. further forming topological maps by self-organization. The useful for studying learning self-organization equation receptive fields. stability analysis shows formation a continuous map emergence block structure responsible columnar microstructures. Kohonen also discussed.