Exchangeability Characterizes Optimality of Sequential Normalized Maximum Likelihood and Bayesian Prediction with Jeffreys Prior
Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, PMLR 22:504-510, 2012.
We study online prediction of individual sequences under logarithmic loss with parametric constant experts. The optimal strategy, normalized maximum likelihood (NML), is computationally demanding and requires the length of the game to be known. We consider two simpler strategies: sequential normalized maximum likelihood (SNML), which computes the NML forecasts at each round as if it were the last round, and Bayesian prediction. Under appropriate conditions, both are known to achieve near-optimal regret. In this paper, we investigate when these strategies are optimal. We show that SNML is optimal iff the joint distribution on sequences defined by SNML is exchangeable. In the case of exponential families, this is equivalent to the optimality of any Bayesian prediction strategy, and the optimal prior is Jeffreys prior.