From the retina to the convolutional network
The hierarchy of cells Hubel and Wiesel find in the cat's visual cortex becomes an artificial network and then convolutional networks.
Hubel, Wiesel, and the visual cortex (October 1959)
Hubel and Wiesel find simple cells in the visual cortex, which respond to an edge in one place, and complex cells, which respond to the same edge over a wider area.
The neocognitron (October 1979)
Fukushima turns that hierarchy into the neocognitron, which alternates layers of S and C cells and recognizes a pattern even when it moves.
LeCun's convolutional network (1989)
LeCun uses backpropagation to train a network with the same architecture of alternating layers.