An artificial neural network simulator with integrated fault simulation capabilities

netFaultSim is a general purpose Artificial Neural Network (ANN) fault simulator capable of specifying arbitrary connectivities, sever al neuron models, and learning. The fault models are manifested as behavioral changes in ANN parameters rather than as defects in component manufacture. The classes of behavioral faults modeled are deletion, stuck, and parameter variation faults. Faults and learning can also be applied to subsets of larger ANNs and several independent regimens may be simultaneously applied. Significantly, a large set of pseudo-randomly generated fault simulations cart be set up and performed from a single invocation of netFaultSim. Example results are presented for autoassociative memory, feedforward, and a large heterogeneous ANN.