Fixed-Horizon Active Hypothesis Testing

Two active hypothesis testing problems are formulated. In these problems, the agent can perform a fixed number of experiments and then decide on one hypotheses or declare its inconclusive. The first problem is an asymmetric formulation in which objective to minimize probability incorrectly declaring particular be true while ensuring that correctly moderately high. This generalization Chernoff–Stein lemma setting. second symmetric misclassification declared conclusively with high probability. For lower upper bounds optimal probabilities derived shown asymptotically tight. Novel experiment selection strategies provided. It using numerical experiments, it demonstrated have significantly better performance nonasymptotic regime.