Asymptotically Best Causal Effect Identification with Multi-Armed Bandits
Date:
Wednesday, November 3, 2021 16:00 - 18:00
Speaker:
Silvia Chiappa (DeepMind London)
Location:
Zoom Link: https://istaustria.zoom.us/j/92619096752?pwd=UWlEYTZ6UGUrSEZWMDNMMDltaEEzUT09 , Meeting ID: 926 1909 6752, Passcode: 189793
Series:
ELLIS talk
Host:
Marco Mondelli
Contact:
Ksenja Harpprecht
In this talk, I will present a method for selecting the causal identification formula with lowest asymptotic variance among a set of available ones. The method assumes an online setting in which the investigator may alter the data collection mechanism in a data-dependent way with the aim of identifying the formula in as few samples as possible, and formalizes this setting using the best-arm-identification bandit framework where the standard goal of learning the arm with the lowest loss is replaced with the goal of learning the arm that will produce the best estimate.