BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Solving Large-scale Machine Learning Problems - Vinod Kumar (IfM-D
 IAL\, University of Cambridge)
DTSTART:20190627T103000Z
DTEND:20190627T113000Z
UID:TALK126568@talks.cam.ac.uk
CONTACT:Shuya Zhong
DESCRIPTION:Big data is one of the major challenges in machine learning\, 
 which leads to slow training and scalability issues of models. In this wor
 k\, we have identified problem formulation\, problem solvers\, optimizatio
 n strategies and platform/framework utilization\, as major areas to tackle
  the challenge. But out of these potential areas\, recently\, researchers 
 have focused on stochastic approximation algorithms\, coordinate descent a
 lgorithms\, proximal algorithms and parallel & distributed algorithms to t
 ackle the challenge.\nWe have utilized the best of stochastic approximatio
 n and coordinate descent approaches to propose a batch block optimization 
 framework (BBOF)\, which has been used with first and second order methods
  to solve the large-scale learning problems. But it has been observed that
  the stochastic approximation and coordinate descent\, do not work well wh
 en combined together because the advantage is lost in extra overhead to im
 plement BBOF due to double sampling\, i.e.\, sampling of data points and t
 hat of features.\nWe have proposed stochastic average adjusted gradient (S
 AAG-I\, II\, III and IV) methods\, as variance reduction techniques to sol
 ve the large-scale problems. We have also proposed stochastic trust region
  Newton (STRON) method\, which solves the Newton system inexactly to handl
 e the large-scale problems. Moreover\, simple sampling techniques have bee
 n proposed to improve the training time of models by reducing the data acc
 ess time. We have provided theoretical analysis and empirical results whic
 h have proved efficacy of proposed methods against the existing techniques
 .
LOCATION:Lecture Theatre 1\, Institute for Manufacturing\, University of C
 ambridge
END:VEVENT
END:VCALENDAR