Dynamically weighted time-delay neural networks for functional approximation

The existing multilayer neural networks are not ideally suited for functional approximation of biological activities as they do not capture the time-related dynamics of biological events. We present a dynamically weighted time-delay neural network (DWTDNN) which is more suited for computations on temporal patterns from biological events. We use this network for modeling the time-series data from fMRI experiments and compare the results with those obtained using existing models of fMRI response.