BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Toward the Design of Efficient and Robust Algorithms for Distribut
 ed Collaborative Deep Learning.
DTSTART:20220110T100000
DTSTAMP:20260407T164015Z
UID:3fe250d6949276c981b26b62ffe25e3262c1840710e8dcfc8e4a1401
CATEGORIES:Conferences - Seminars
DESCRIPTION:Tao Lin\, doctoral student in Machine Learning and Optimizatio
 n Laboratory\, EPFL\nThe EDIC program is happy to invite you to the public
  talk by our doctoral student Tao Lin.\nThe aim of the talk is to present 
 his achievements to a broad audience to prepare for hiring interviews comi
 ng up soon. Be sure to join\, listen to the talk and participate in the Q&
 A session at the end of the presentation.\n\nAbstract:\nRecent progress in
  deep learning has shown impressive results in many domains\, which has al
 so led to a dramatic increase in the size\, complexity\, and computational
  power of the training system. In the meantime\, significant efficiency co
 ncerns and privacy risks arise as currently most learning is operated in a
  centralized way\, while tremendous amounts of data tend to be generated o
 n decentralized edge devices and may contain users’ personal information
 . All these motivate the necessity of moving to distributed deep learning.
 \nIn this talk\, I will highlight my three key steps toward enabling distr
 ibuted collaborative deep learning. I will show how to improve the learnin
 g efficiency of distributed deep learning in data centers\, for both commu
 nication-restricted settings and the regime of poor generalization of larg
 e-batch SGD. In addition to learning in the data center\, I will also disc
 uss my attempts at improving the efficiency of learning on edge devices\, 
 where the advances come from the decentralized and sparse communication to
 pology. Finally\, I will discuss the demands of deploying robust decentra
 lized learning algorithms\, due to the hampered learning performance induc
 ed by heterogeneous learning environments. I will then conclude the talk 
 with an outlook on future directions.\n\nBio:\nTao Lin is a fifth-year Ph
 .D. student at École Polytechnique Fédérale de Lausanne (EPFL)\, superv
 ised by Prof. Martin Jaggi and Prof. Babak Falsafi. Prior to that\, he rec
 eived a Master of Science degree from EPFL and a Bachelor of Engineering d
 egree from Zhejiang University (ZJU). His recent research interests includ
 e designing efficient and robust optimization algorithms for distributed c
 ollaborative deep learning.\n 
LOCATION:https://epfl.zoom.us/j/69449724565
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR