Linear Filter Model of Synapse Operating on Noisy Inputs: Associated Neural Networks

摘要

In this research paper, based on biological motivation, synapses are modeled as frequency selective filters. These filters remove undesired noise signals. Also, a synapse is modeled as a matched filter which maximizes the signal-to-noise ratio at the neuron activation level. Utilizing the results in optimal linear filtering, the synapses are modeled as Minimum Mean Square Error (MMSE) Linear Filters. It is hoped that these models provide robust neuronal information processing in the presence of noise signals.