Under model misspecification, Thompson Sampling's posterior still converges exponentially to a pseudo-truth set, but within that set multiple parameters can persist and the MAP estimate may not converge.
Which misspecifications persist?
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
econ.EM 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Dynamic Decision-Making under Model Misspecification
Under model misspecification, Thompson Sampling's posterior still converges exponentially to a pseudo-truth set, but within that set multiple parameters can persist and the MAP estimate may not converge.