Found 1 document(s)
Page 1 of 1
-
Classical inverse bandit methods typically aim to recover true reward parameters. In contrast, we infer a demonstrator's subjective beliefs, specifically its internal estimates of suboptimality gaps. We consider no-regre...