TU4.R4.2

Sequential Adversarial Hypothesis Testing

Eeshan Modak, Tata Institute of Fundamental Research, India; Mayank Bakshi, Arizona State University, United States; Bikash Kumar Dey, Indian Institute of Technology, Mumbai, India; Vinod M. Prabhakaran, Tata Institute of Fundamental Research, India

Session:
Sequential Hypothesis Testing and Change Detection

Track:
11: Information Theory and Statistics

Location:
Omikron II

Presentation Time:
Tue, 9 Jul, 16:25 - 16:45

Session Chair:
I-Hsiang Wang, National Taiwan University
Abstract
We study the adversarial binary hypothesis testing problem \cite{brandao2020adversarial} in the sequential setting. Associated with each hypothesis is a closed, convex set of distributions. Given the hypothesis, each observation is generated according to a distribution chosen (from the set associated with the hypothesis) by an adversary who has access to past observations. In the sequential setting, the number of observations the detector uses to arrive at a decision is variable; however there is a constraint on the expected number of observations used. We characterize the closure of the set of achievable pairs of error exponents.
Resources