Probability in Amsterdam
Posted by David Corfield
I’m off to Amsterdam today to attend the FotFS VI conference. I’m going to be speaking on what I’ve learned from my time amongst machine learning researchers (slides). It occurred to me that my entry into statistical learning theory has been slowed by its largely belonging to the culture of mathematics with which I am less familiar. E.g., the PAC theorem mentioned on slide 5 (page 4) derives from the ‘combinatorics’ culture, see also this related guest post on Tao’s blog. Perhaps this explains why I keep finding myself wanting to find more geometry about the place, as information geometry offers.
By the way, while I’m mentioning Tao’s blog, take a look at his post on a Fields Medalist Symposium held at UCLA. Café members may be especially interested in Tao’s descriptions of Vaughan Jones’ and Richard Borcherd’s talks. Do we have anything to say here about Borcherds’ Levels 0 to 4?
Posted at May 1, 2007 8:44 AM UTC
QFT, Feynman, I link to Branko Malesevic; Re: Probability in Amsterdam
About Borcherds’ Levels 0 to 4, I’d note that there is a combinatorial structure to the number of meaningful differential operation compositions on the space R^n. Do those emerge easily from an n-categorical treatment?
I link to Branko Malesevic’s paper on that through the OEIS, and give a table:
http://www.research.att.com/~njas/sequences/?q=A116183&language=english&go=Search
The QFT talk is interesting, but sweeps under the rug the difficulties with putting Feynman path integrals on a sound footing – though nobody disputes his boldness and creative genius in inventing them semi-intutitively.