Measurement of nonlinear 2nd-order kernels using Gaussian and natural inputs

We implemented and investigated two frequency-domain based methods for measuring the nonlinear kernels of homogeneous 2nd-order 2D Volterra systems. For Volterra systems excited by Gaussian noise, there exists a closed-form solution L. Tick (1961). In the present work, however, we are interested in measuring the kernels of visual cortical neurons, which empirically show only weak responses to white Gaussian noise. Hence, we also implemented a technique based on an algorithm developed in K. Kim et al., (1988) that measures kernels with generalized inputs. We evaluate the results regarding noise-stability and convergence properties using Gaussian and naturalistic stimuli for nonlinear models of cortical cells.