The fifth Theory Weekly Meeting will happen this Friday between 4:00 – 5:00 pm at 203 Davis Hall(Theory Lab). The schedule will be as follows :
4:00 – 4:50 pm – A talk by Atri Rudra. The abstract of the talk in his own words –
A non-divide and conquer FFT
I will describe a description of the FFT (Fast Fourier Transform) that does not use the textbook Divide and Conquer paradigm. This is a message passing algorithm for the FFT and it has the following “advantages” over the standard description: (i) (In my personal opinion) this version makes it more transparent as to why we can get the O(nlog(n)) runtime for FFT when compared to the naive O(n^2) runtime; and (ii) This version also generalizes to show a class of matrices (of which the DFT is a special case) for which matrix vector multiplication can be done in near linear time.
Just to clarify: this is a *known* description that (in my informal polls) seems to be not that well-known. Also for simplicity I’m going to show a sub-quadratic version of the FFT.
4:50 – 5:00 pm – Socializing with the Theory community.
All are welcome!