Artificial Consciousness/Neural Correlates/Neural Network Models/Hidden Layer
Hidden Layer/s[edit | edit source]
One of the problems with the Hebbian and HH Neuron models was that they didn't capture the complexity of the dendritic mass well. In order to compensate, it was suggested that a Hidden Layer be incorporated into the neuron, to cover the dendritic mass. In some cases this had to be expanded to multiple layers because the dendritic mass was more highly interconnected than at first thought.
This model is based on the idea that any branching in the dendrite, can be modeled as if it were a neuron, and the complexity of all the branches, can be modeled as a network with limited connectivity. By replacing the separate small networks with a single larger network, you can achieve the same processing with a simpler model. This abstraction is based on the discovery that fully interconnected networks required fewer hidden neurons to achieve the same process. A significant amount of research went into determining the optimal number of neurons for the hidden layers, and some network models automatically adjust for the processing requirements so that the actual number of hidden neurons is not needed to be known in order to simulate the same process.