How to Get Rid of Noise in a Correlation Matrix?

MRI is ultimately about separating a known but variable signal from highly variable noise. How one does this makes all the difference. fMRI is particularly challenging since what is signal and what is noise is not always clear, as they both vary in time and space. In this episode, Peter talks to Jack Gallant, PhD, a neuroscientist and engineer. Jack is currently a Chancellor’s Professor of Psychology and Class of 1940 Endowed Chair at UC Berkeley and is affiliated with the Department of Electrical Engineering and Computer Science. He is a huge proponent of fMRI encoding or, more generally, careful model building to probe the time series. He thinks that more model free approaches and paradigm free methods are ultimately limited. The discussion gets technical as well as intense at times; while Jack and Peter agreed most of the time, there were some nuanced differences of opinion - mostly when it came to discussing alternative methods for probing fMRI data. Overall, we think it was a fun and hopefully a useful discussion!  What comes through is Jack’s passion for what he does. Given that they only barely got started with Peter’s questions, Peter invited him back for another chat - see S2 Episode 6!