Artificial Consciousness/Neural Correlates/Neural Models/Hebb Neuron

From Wikiversity
Jump to navigation Jump to search

The Hebb Neuron (Neural Network Model)[edit | edit source]

Although Hebb was by no means the discoverer of the Neuron, nor even the first scientist to attempt to model it, his model was easily understood, and translated well to computer programs, which made it easily used to simulate neural networks. A simple adjusted aggregation of synapses, the calculation translated out to a Weighted sum adjusted to fit the neural logarithmic curve, minus a threshold value that indicated neural sensitivity.

σΣj-t


In his book Perceptron Marvin Minsky however leveled the devastating criticism that Hebbs Neuron did not adequately capture second order characteristics detected in natural neurons. Neural Network theory thereafter died its first death, and Neural Network Studies have since gone through a number of periods of popularity interspersed with temporary abandonment, only to be picked up again as newer better models were discovered.

The main problem with Neural Networks is simply we don't know enough to adequately model the neuron, so our models tend to be oversimplified, and easily shot down, when compared against the natural networks. It is important to note that Neural Networks are Simulations of Neural systems, and cannot be expected to adequately model them at this early stage in their development. The capability of the network models to capture some of the function of the neuron in its network, is what keeps attracting scientists back to them, no matter how many times they have been disappointed in the models capability.