BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning Markov Networks for Mixed Big Data: Applications to Cance
 r Genomics - Genevera Allen\, Rice University
DTSTART:20150305T160000Z
DTEND:20150305T170000Z
UID:TALK57846@talks.cam.ac.uk
CONTACT:20082
DESCRIPTION:"Mixed Data'' comprising a large number of heterogeneous varia
 bles (e.g.  count\, binary\,\ncontinuous\, skewed continuous\, among other
  data types) is prevalent in varied areas such as\nimaging genetics\, nati
 onal security\, social networking\, Internet advertising\, and our particu
 lar\nmotivation - high-throughput integrative genomics.  There have been l
 imited efforts at\nstatistically modeling such mixed data jointly\, in par
 t because of the lack of computationally\namenable multivariate distributi
 ons that can capture direct dependencies between variables of\ndifferent t
 ypes.  \nIn this talk\, we address this by introducing several new classes
  of Markov Random Fields (MRFs)\,\nor graphical models\, that yield joint 
 densities over mixed variables. To begin\, we present a\nnovel class of MR
 Fs arising when all node-conditional distributions follow univariate\nexpo
 nential family distributions that\, for instance\, yield novel Poisson gra
 phical models. \nNext\, we introduce extensions of this for Mixed MRF dist
 ributions.  Unfortunately\, these\nformulations can place severe and unrea
 listic restrictions on the parameter space.  To remedy\nthis\, we we intro
 duce a class of mixed conditional random field distributions\, that are th
 en\nchained according to a block-directed acyclic graph to form a new clas
 s of so-called Block\nDirected Markov Random Fields (BDMRFs). The Markov i
 ndependence graph structure underlying our\nBDMRF then has both directed a
 nd undirected edges. \n\nWe will briefly review the theoretical properties
  of these models and introduce penalized\nconditional likelihood estimator
 s with statistical guarantees for learning the underlying mixed\nnetwork s
 tructure. Simulations as well as an application to integrative cancer geno
 mics\ndemonstrate the versatility of our methods.  In our particular examp
 le\, we learn integrative\ngenomic networks from breast cancer next genera
 tion sequencing expression data and mutation data\nthat yield several inte
 resting findings.\n \nJoint work with Eunho Yang\, Pradeep Raviukmar\, Zha
 ndong Liu\, Yulia Baker\, and Ying-Wooi Wan.
LOCATION:MR12\,  Centre for Mathematical Sciences\, Wilberforce Road\, Cam
 bridge
END:VEVENT
END:VCALENDAR