Statistical Risk Models, Billion Alphas, and Cancer Signatures | Hanlon Financial Systems Center

Statistical Risk Models, Billion Alphas, and Cancer Signatures

Statistical Risk Models, Billion Alphas, and Cancer Signatures

seminar date: 
Wednesday, May 4, 2016 - 6:45pm
seminar location: 
BC541
Zura Kakushadze President & Co-Owner, Quantigic Solutions; Full Professor, Free University of Tbilisi
Abstract: 

We discuss how to construct statistical risk models, including methods for fixing the number of risk factors. One such method is based on eRank (effective rank). One application is computing optimal weights for combining a large number N of alphas. The algorithm does not cost O(N^3) or O(N^2) operations but is much cheaper: optimization simplifies when N is large and the # of operations scales as N. We also present a novel method for extracting cancer signatures from genome data, including identification and removal of somatic mutational noise. We apply nonnegative matrix factorization (NMF) to 1389 genome samples aggregated by 14 cancer types and filtered using our method. The resultant cancer signatures have substantially lower variability than from raw data. The computational cost is cut by a factor ~10. We find 3 novel cancer signatures, including a liver cancer dominant signature (96%) and a renal cell carcinoma signature (70%). 

 

Bio: 

Zura Kakushadze received his Ph.D. in theoretical physics from Cornell University at 23, was a Postdoctoral Fellow at Harvard University, and an Assistant Professor at the C.N. Yang Institute for Theoretical Physics at Stony Brook. Dr. Kakushadze received the Alfred P. Sloan Fellowship in 2001.  After expanding into quantitative finance, he was a Director at RBC Capital Markets, Managing Director at WorldQuant, Executive Vice President and substantial shareholder at Revere Data, Adjunct Professor at UConn, and currently is the President and co-owner of Quantigic Solutions and a Full Professor at Free University of Tbilisi.  He has 110+ publications in physics, finance and other fields, 3,300+ citations and h-index 31 (physics), 41,000+ downloads and top-10 twelve-month rank on SSRN (finance), and 200,000+ followers on LinkedIn.

The Financial Engineering Seminar Series is a centerpiece of the graduate Financial Engineering program.  Its mandate is to arrange talks on current research and industry trends in financial engineering and quantitative finance that will be of interest to those who work in both industry and academia. This event is sponsored by the School of Systems & Enterprises, Financial Engineering Division.