An Efficient Bayes Coding Algorithm for Changing Context Tree Model

The context tree model has the property that occurrence probability of symbols is determined from a finite past sequence and broader class sources includes i.i.d. or Markov sources. This paper proposes non-stationary source with models change interval to interval. Bayes code for this requires weighting posterior probabilities points, so computational complexity it usually increases exponential order. Therefore, challenge how reduce complexity. In paper, we propose special prior distribution points develop an efficient coding algorithm by combining two existing algorithms. minimizes risk function proposed in polynomial We investigate behavior performance conducting experiments.