Inferring Neural Firing Rates from Spike Trains Using Gaussian Processes
JP Cunningham, BM Yu, KV Shenoy - Stanford University, USA M Sahani - Gatsby Computational Neuroscience Unit, UCL, UK

Poster W66

Method:
·GP regression framework with inference and model selection ·Firing rate prior: ·Inhomogeneous Gamma interval process observation model:

Spike train GP estimate True rate Other methods

For the Neuroscience Community:
·Automatic smoothness detection ·Error bars on firing rate estimates ·Single trial estimates ·Performance improvements vs. conventional smoothers

For the Machine Learning Community:
·GP for point process intensity estimation ·Solving a nonnegative constrained inference on a non-zero mean GP without a link function · Numerical techniques for computational tractability