Latent Gaussian Process for Spike Data


While I’m no longer able to participate in the Statistical Learning Reading Group at Michigan, I was a huge fan of reading the articles each week and thought I would post the slides of the articles I presented here in case anyone happens to find themselves interested.
While I don’t mention it in these slides specifically, I remember there being a widespread concern about the likely non-identifiability of the recovered parameters in this model, a common issue in machine learning methods. While this wouldn’t be as big of a concern if the aim were prediction, the authors did have the inference of the latent gaussian process as one of their inferential goals. Still, a very stimulating paper overall. It was originally presented at NIPS in 2018. The slides are linked here. and the pdf can be found here.

Adam Peterson
Graduate Research Assistant

Adam Peterson is Biostatistics PhD candidate at the University of Michigan.