Dear Stan community:
Could you guys, please, suggest any topics for a project (bachelor or master level) that you consider attractive, and in which gaussian process could be applied.
Regards.
GPs on a sphere+time for use in Neuroimaging (EEG, MEG, fNIRS) would be neat.
1 Like
Or, if you don’t want to fuss with the (so far as I understand) unsolved issue of GPs on a sphere, focus on fMRI data where you are dealing with a nice Cartesian space throughout the head. Has the “benefit” of not being as highly resolved in time too, which makes for less intensive computation. There’s a paper whose title starts “cluster failure” on the unsupported assumptions of more traditional fMRI smoothing/cluster analysis and to which I’ve always thought GPs would be an elegant solution
2 Likes
Honestly it’d be a better model for COVID modeling than a third order polynomial.
2 Likes
I think the mechanisms behind disease transmission are too complex for GP, not sure tho.
Yes, I believe @saudiwin was making a joke at the expense of the “cubic model” being widely mocked on social media in the last couple days.
2 Likes
Sorry probably not the best forum for sarcasm.
More constructively, you should look in to Twitter data. GPs are very useful with irregular time series because they don’t require regularly-placed intervals, and Twitter content has very precise and irregular time markers.
2 Likes