In Predictive Systems Lab we look at artificial neural networks exactly as described elsewhere in the literature, except we make a distinction regarding the ownership of neural connection weights. Some authors prefer to specify weights as the property of the firing layer; we, on the other hand, prefer to let weights be the property of the receiving layer (see the figure below).

 

 

We define the output of any one ith neuron in layer L, L > input layer, as

 

Where is the activation function, the output of the jth neuron of layer L-1, the number of neurons in layer L-1, the weight of this neuron for , and the threshold or bias weight from a constant node whose output is always 1.