BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:"Design and Analysis of Scalable Algorithms via Statistical Tools"
DTSTART:20170208T140000
DTEND:20170208T150000
DTSTAMP:20260925T120633Z
UID:bda8b99c49bdb8002de3738e2e46e333e83cffe44acacee4f6e7f2b8
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Murat A. Erdogdu (Stanford University) \nStatistics and op
 timization have been closely linked since the very outset. This connection
  has become more essential lately\, mainly because of the recent advances 
 in computational resources\, the availability of large amount of data\, an
 d the consequent growing interest in statistical and machine learning algo
 rithms.\n \nIn this talk\, I will discuss how one can use tools from stat
 istics such as Stein’s lemma and subsampling to design scalable\, effici
 ent\, and reliable optimization algorithms. The focus will be on large-sca
 le problems where the iterative minimization of the empirical risk is comp
 utationally intractable\, i.e.\, the number of observations n is much larg
 er than the dimension of the parameter p\, n >> p >> 1. The proposed algor
 ithms have wide applicability to many supervised learning problems such as
  binary classification with smooth surrogate losses\, generalized linear p
 roblems in their canonical representation\, and M-estimators. The algorith
 ms rely on iterations that are constructed by Stein’s lemma\, that achie
 ve quadratic convergence rate\, and that are cheaper than any batch optimi
 zation method by at least a factor of O(p).\n \nI will discuss theoretica
 l guarantees of the proposed algorithms\, along with their convergence beh
 avior in terms of data dimensions. Finally\, I will demonstrate their perf
 ormance on well-known classification and regression problems\, through ext
 ensive numerical studies on large-scale real datasets\, and show that they
  achieve the highest performance compared to other widely used and special
 ized algorithms.\n \n 
LOCATION:CIB lecture hall : BI A0 448 http://plan.epfl.ch/?room=BIA0448
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR