Neural Nets - History and People

Warren McCulloch and Walter Pitts, Marvin Minsky, Paul Werbos and John Hopfield

Should it be surprising that Alan Turing foresaw the development of artificial neural networks?

In 1943 Warren McCulloch and Walter Pitts published the pioneering article "A Logical Calculus of Ideas Immanent in Nervous Activity" in which they proposed the first artificial neural network based solely on algorithms and mathematics. A commentary on their paper is here.

In 1951 Marvin Minsky built the SNARC, the first neural network simulator. In 1969, Minsky and Seymour Papert (the inventor of LOGO) published the book Perceptrons, which laid the foundations for modern artificial neural nets.

The first hardware neural net was built by Frank Rosenblatt in 1960. He pioneered the concept of training a neural net by changing the strength of its synaptic signals.

In 1974, Paul Werbos introduced the back propagation algorithm, which allows errors in a neural network to propagate back to previous neuron layers. This dramatically speeds up training the network.

In 1982, John Hopfield invented the associative neural network--called the Hopfield net. His pioneering paper is here.

Because the development of neural networks has taken place relatively recently and because it covers such a large area, there are many other noted researchers in the field.