BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/961
DTSTAMP:20230914T125945Z
SUMMARY:Recovering Causality Graphs from Time Series Data
DESCRIPTION:Speaker: Ravi R. Mazumdar (University Research Chair Professor\
nDept. of Electrical and Computer Engineering\nUniversity of Waterloo\, Ca
nada)\n\nAbstract: \nAbstract: Suppose we have N time series available whe
re one time-series could be causally dependent on others. For example\, su
ch dependence can be found in economic data or weather data. The goal is t
o recover the directed causality graph that links these time series.\nAs i
s well known causality and correlation are not the same and thus one of th
e important questions is how to address this issue. There are several fram
eworks such as directed information\, the notion of Granger causality\, et
c. However working with directed information requires too much a priori kn
owledge about the structure of the time series that is unavailable.\nIn th
is talk I will show how the notion of Granger causality can be tied to Wie
ner filtering that allows us to recover a directed random graph whose edge
s are represented by the innovations filters. This approach as well as the
directed information approach assuming Gaussianity are however quite comp
utationally intensive. To address this issue we show how it is possible to
consider a sparse problem based on a mixed L1 - H1 norm\, a generalized G
LASSO approach that takes into account the temporal dependencies that lead
s to an approach for selecting edges in a directed graph characterized by
complex polynomials. This results in a convex optimization problem and mod
ern techniques such as ADMM are well suited to such problems.\nWe then dis
cuss the sparsification problem associated with Granger causality graphs w
hen local neighbourhoods are used. The basic issue is to find the sub-grap
h of the causality graph that is closest to the original in terms of a sui
table norm. In this context we study both the l2 as well as the graph supp
ort recovery problem subject to sparsity constraints [jointly with Syamant
ak Dattagupta (Morgan Stanley) and Ryan Kinnear (Waterloo)].\nBio: The spe
aker was educated at the Indian Institute of Technology\, Bombay (B.Tech\,
1977)\, Imperial College\, London (MSc\, DIC\, 1978) and obtained his PhD
under A. V. Balakrishnan at UCLA in 1983.\nHe is currently a University R
esearch Chair Professor in the Dept. of ECE at the University of Waterloo\
, Ont.\, Canada where he has been since September 2004. Prior to this he w
as Professor of ECE at Purdue University\, West Lafayette\, USA. He is a D
.J. Gandhi Distinguished Visiting Professor at the Indian Institute of Tec
hnology\, Bombay. He is a Fellow of the IEEE and the Royal Statistical Soc
iety. He is a recipient of the Best Paper Awards at INFOCOM 2006\, the Int
ernational Teletraffic Congress 2015\, Performance 2015\, and was runner-u
p for the Best Paper Award at INFOCOM 1998.\nHis research interests are in
complex networks\, stochastic analysis\, and randomized algorithms.\n
URL:https://www.tcs.tifr.res.in/web/events/961
DTSTART;TZID=Asia/Kolkata:20190426T160000
DTEND;TZID=Asia/Kolkata:20190426T170000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR