Predictive stochastic complexity and model estimation for finite-state processes

摘要: Abstract It is shown that the predictive and nonpredictive stochastic complexities relative to class of finite-state models are asymptotically equivalent in a probabilistic sense. To this end, universal, sequential, noiseless coding scheme attaining minimum description length (MDL) data with respect presented investigated. relies on an MDL-based estimator model structure, which proved be strongly consistent. An interpretation result process ‘close’ every class, regardless can constructed. This universal employed solution sequential decision problems like coding, prediction, gambling, optimal manner.