Loading timeline…
19801980s
Neocognitron
Kunihiko Fukushima's hierarchical vision network — the convolutional architecture, a decade before it had the name.
Why It Was Important
The Neocognitron introduced alternating layers of local feature detectors and pooling units, giving a network tolerance to shifts in position. That is precisely the design LeCun's LeNet would later train with backpropagation and AlexNet would scale onto GPUs. Modelled on Hubel and Wiesel's simple and complex cells, it is the most direct ancestor of every convolutional network in use today.
Who Invented It
Kunihiko Fukushima
Japanese computer scientist at the NHK research laboratories.
Applications
- Handwritten Character Recognition
- Pattern Recognition
- Computer Vision
Key Papers
- Neocognitron: A Self-Organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in Position
Kunihiko Fukushima · Biological Cybernetics · 1980