Select another
Cluster

Lecture: Don't Bet on your Random Number Generator

Date: 31 March 2017

Place: Room E004/005 (Ground floor), JFK Building, 29, avenue John F. Kennedy, L-1855

Abstract

Pseudo-random numbers are used in countless contexts, including jury selection, electronic casino games, physical and chemical simulations, numerical integration, random sampling, Monte Carlo methods, stochastic optimization, and cryptography. They are used in scientific fields from sociology to particle physics. It's tempting to think that the pseudo-random number generators (PRNGs) in software packages and programming languages in common use are "good enough" for most purposes. Resist the temptation. Pigeonhole arguments and empirical results show that PRNGs in statistical software packages and general purpose programming languages are not adequate for basic statistical purposes such as random sampling, generating random permutations, and the bootstrap – even for relatively small data sets. Cryptographers have developed cryptographically secure pseudo-random number generators (CS-PRNGs), which provide a far better approximation to truly random numbers, as manufacturers of gambling machines are well aware. Statistical packages and general-purpose programming languages should use CS-PRNGs by default.

Philip B. Stark is Professor of Statistics and Associate Dean of Mathematical and Physical Sciences at the University of California, Berkeley. He works on inference and uncertainty quantification in applications including astrophysics, earthquakes, ecology, elections, gender bias, legislation and litigation, particle physics, public health, and public policy. He received a Presidential Young Investigator Award, the John Gideon Award for Election Integrity, the UC Berkeley Chancellor's Award for Research in the Public Interest, and the Leamer-Rosenthal Prize for Transparency in Social Science. Stark currently serves on the Board of Advisors of the US Election Administration Commission. He holds an AB from Princeton University and a PhD from the University of California, San Diego.

Back to all events

More info

To register, please email  snt-events@uni.lu