neocognitron

A visual pattern recognizer built as a stack of neural layers, designed by Kunihiko FukushimaKunihiko Fukushima, prop. n.: A Japanese engineer who in 1979 invented the neocognitron, a brain-inspired image-recognition network that was the direct ancestor of today's convolutional neural networks. in 1979–1980 and the direct ancestor of the convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects.. It borrows the two cell types that David Hubel and Torsten Wiesel described in the visual cortex: its S-layers apply one set of feature detectors at every position of the image, and each following C-layer combines the responses of neighboring positions, so a feature still registers when it is shifted or slightly distorted. Stacking these stages yields detectors for progressively larger and more abstract patterns, so that a character is recognized wherever it appears. The neocognitron learned layer by layer through unsupervised, competitive self-organization (later versions added some supervision) rather than end-to-end backpropagationbackpropagation, n.: The method neural networks use to learn from mistakes: it works backward from the error to figure out how much each internal connection contributed, so every connection can be adjusted in the right direction., which LeNetLeNet, prop. n.: A family of early image-recognition neural networks built at Bell Labs from 1989 that learned to read handwritten digits. A version read huge numbers of bank checks, and its design underlies today's image models. later supplied.

The convolution and pooling layers of a modern CNN echo the S-cells and C-cells of the neocognitron.