Topic: Measuring the change in output signal according to input change in neuronal networks.

Abstract:
Computer simulation of native neuronal networks provides a way to controlsignal propagation through those networks. Given the topology of a neuronal network, an input node which is assigned some particular sequence of binary signals and an output node, we measure the signal sequence received at the output node. The computer program has been developed in order to perform this kind of simulations. However, another important question still remains open: is there any correlation between the extent of input signal change and the extent of output signal change. Taking into consideration that some particular small changes in input signals may cause unpredictable sharp twists in the output, we predict that on average there is still straight correlation between the extent of input change and the extent of output change. That is, the smaller the change in input, the smaller the change in output and, inversely, the greater the change in input, the greater the change in output. This hypothesis is put to the test. We use a sample neuronal network to run simulations with different sets of inputs partitioned according to the number of input bits changed and measure the variance of the change in resulted outputs.