Loading Events

« All Events

  • This event has passed.

OR Seminar: Raghu Pasupathy

February 20, 2023 @ 4:30 pm - 5:45 pm

FREE
Raghu Pasupathy | Seminar Series

Join us in welcoming Raghu Pasupathy, professor of statistics at Purdue University, as he discusses his research.

Missed the Presentation?

Watch the recording NOW
Password: p22CSfD$

Title

Batching as a General Statistical Inference Device

Abstract

Consider the following example problem settings: (i) estimate the 90th percentile time-dependent inventory level at multiple points in a large simulated supply chain; (ii) estimate the optimal glass cooling schedule as the solution to a stochastic optimization problem subject to PDE constraints relating to heat flow; and (iii) estimate the time-dependent expected number of infections in an epidemic that is modeled using an SDE. Each of (i) — (iii) constitutes a setting where a time series of data obtained through some means, e.g., a simulation, is “processed” in the service of estimating a parameter such as a percentile curve or the solution to an optimization problem. We consider statistical inference in such contexts, whereby one seeks to quantify the error in an obtained estimator, e.g., through a confidence region or a hypothesis test.

Historically, inference within complicated computational contexts has been considered challenging because the parameter needing estimation is often not a mean, and the input time series is non-normal and exhibits heavy dependence. We argue, however, that the remarkably simple idea of batching might provide a solution. Batching, like the bootstrap, is a resampling idea and works in three steps: (i) divide the input time series into overlapping batches; (ii) construct parameter estimates from each batch; and (iii) appropriately use the batch estimates after accounting for dependence, to perform statistical inference. The resulting procedures are usually trivial to implement in software and, as we show, are provably correct and efficient. Batching ideas originated in the 1950s and have enjoyed steady development in the simulation community since the 1970s, mostly within the problem of variance estimation. Our thesis is that batching ideas have much wider utility. Time permitting, I will discuss batching’s relationship to the bootstrap and subsampling, along with numerical examples.

Biography

Raghu Pasupathy is a Professor of Statistics at Purdue University. Prior to joining Purdue in 2014, Paspathy spent nine years in the Industrial and Systems Engineering Department at Virginia Tech, first as an assistant professor and then as an associate
professor. Pasupathy’s research interests lie in the theoretical and computational aspects of stochastic optimization. Pasupathy has been associated with the simulation and optimization communities in various capacities over the previous two decades, including serving as President of the INFORMS Simulation Society from 2018 — 2020 and
as an editor for ACM TOMACS, Operations Research, INFORMS Journal on Computing, IISE Transactions, and Mathematical Programming. More information, including downloadable papers and computational software, can be obtained through his website at https://web.ics.purdue.edu/~pasupath/

Details

Date:
February 20, 2023
Time:
4:30 pm - 5:45 pm
Cost:
FREE
Event Category:
Event Tags:
, , ,

Venue

4290 Fitts-Woolard Hall
915 Partners Way, Room 4290
Raleigh, NC 27606 United States
+ Google Map