3
The xyz algorithm for fast interaction search in highdimensional data
http://sms.cam.ac.uk/media/2649186
Shah, R
Thursday 18th January 2018  16:00 to 16:45
40
The xyz algorithm for fast interaction search in highdimensional data
http://sms.cam.ac.uk/media/2649186
http://rss.sms.cam.ac.uk/itunesimage/2647863.jpg
no

The xyz algorithm for fast interaction search in highdimensional data
ucs_sms_2647052_2649186
http://sms.cam.ac.uk/media/2649186
The xyz algorithm for fast interaction search in highdimensional data
Shah, R
Thursday 18th January 2018  16:00 to 16:45
Fri, 19 Jan 2018 15:43:38 +0000
Isaac Newton Institute
Shah, R
b0694fea5e3a069d8caa2b2d8147ec47
b7878bd5b5d3d46453bed09b99c10a48
9930a4d799eb1559f4d13f4486fdd435
d05d3f6f051fbbc950e8d9b639a4d626
Shah, R
Thursday 18th January 2018  16:00 to 16:45
Shah, R
Thursday 18th January 2018  16:00 to 16:45
Cambridge University
2773
http://sms.cam.ac.uk/media/2649186
The xyz algorithm for fast interaction search in highdimensional data
Shah, R
Thursday 18th January 2018  16:00 to 16:45
When performing regression on a dataset with pp variables, it is often of interest to go beyond using main effects and include interactions as products between individual variables. However, if the number of variables pp is large, as is common in genomic datasets, the computational cost of searching through O(p2)O(p2) interactions can be prohibitive. In this talk I will introduce a new randomised algorithm called xyz that is able to discover interactions with high probability and under mild conditions has a runtime that is subquadratic in pp. The underlying idea is to transform interaction search into a much simpler close pairs of points problem. We will see how strong interactions can be discovered in almost linear time, whilst finding weaker interactions requires O(pu)O(pu) operations for 1<u<21<u<2 depending on their strength. An application of xyz to a genomewide association study shows how more than 10111011 interactions can be screened in minutes using a standard laptop. This is joint work with Gian Thanei and Nicolai Meinshausen (ETH Zurich).
20180119T15:43:38+00:00
2773
2649186
true
16x9
false
no