Thesis ideas

I’ve been kicking around some different ideas for my Master’s thesis recently. I want to do something involving machine learning the temporal relationships between neuron spike trains. That’s about as far as I’ve gotten.

I recently read a paper by Brillinger about graphical models of point processes. Learning one of these seems like an interesting option.

On the other hand, I’m kind of enamored with learning Haskell right now. Maybe I can do a stream fusion online variational bayes learning of point processe graphical model structures? Seems like a bit of a mouthful.. ;)

This blog is probably going to focus largely on the statistics as I get through learning it, with hopefully minor bits of neuroscience if I come across them. Currently I’m working through The Bayesian Choice by Christian P. Robert, and I have to admit it’s somewhat heavy going.

A charming quip in the preface explains that in this second edition, exercises were added to form a /mix/ of difficulty levels, rather than the “universally hard” selection of the previous editions. But when you’re lying flat on your back, a 5-foot shrimp and a 6-foot bruiser are both pretty tall.