# Neocognitron **Domain:** Neural Networks / Computer Vision **Doc Type:** Historical Architecture Node **Maturity:** Foundational ## Definition The **neocognitron** is a multilayer self-organizing recognition architecture developed by [[wiki/Kunihiko Fukushima|Kunihiko Fukushima]]. It alternates plastic S-cells with pooling C-cells to build representations tolerant of changes in stimulus position. ## Historical Context Fukushima presented the architecture at the 1979 International Joint Conference on Artificial Intelligence and published an expanded account in 1980. ## Architectural Legacy The neocognitron is a direct ancestor of the [[wiki/Convolutional Neural Network|convolutional neural network]] because it combines local receptive fields, layered feature extraction and spatial pooling. ## See Also [[wiki/Cognitron|Cognitron]], [[wiki/Artificial Neural Networks|Artificial Neural Networks]], [[wiki/Pattern Recognition Systems|Pattern Recognition Systems]]