Computer Science – Learning
Scientific paper
2004-07-16
Proc. 15th International Conf. on Algorithmic Learning Theory (ALT-2004), pages 294-308
Computer Science
Learning
17 pages
Scientific paper
We consider the Minimum Description Length principle for online sequence prediction. If the underlying model class is discrete, then the total expected square loss is a particularly interesting performance measure: (a) this quantity is bounded, implying convergence with probability one, and (b) it additionally specifies a `rate of convergence'. Generally, for MDL only exponential loss bounds hold, as opposed to the linear bounds for a Bayes mixture. We show that this is even the case if the model class contains only Bernoulli distributions. We derive a new upper bound on the prediction error for countable Bernoulli classes. This implies a small bound (comparable to the one for Bayes mixtures) for certain important model classes. The results apply to many Machine Learning tasks including classification and hypothesis testing. We provide arguments that our theorems generalize to countable classes of i.i.d. models.
Hutter Marcus
Poland Jan
No associations
LandOfFree
On the Convergence Speed of MDL Predictions for Bernoulli Sequences does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with On the Convergence Speed of MDL Predictions for Bernoulli Sequences, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and On the Convergence Speed of MDL Predictions for Bernoulli Sequences will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-636282