Non-Linear Neural Spike Train Decoding via Polynomial Kernel Regression

Neural spike train decoding is an important task for understanding how the biological nervous system performs computation and communication. Interest in this field has grown due to recent advances in neural prosthetics, as well as the need to explore non-traditional computational architectures. Current methods deal poorly with or ignore altogether the non-linearities inherent in neural computation. In this paper, we explore polynomial kernel regression as a method for for dealing with neural non-linearities. Two experiments, based on sensory perception and motor control, demonstrate the ability of the approach to decode the neural spike code using synthetic data.