In this paper we analyze the average off-training-set behavior of the Bayes-optimal and Gibbs learning algorithms. We do this by exploiting the concept of refinement, which concerns the relationship between probability distributions. For non-uniform sampling distributions the expected off-training-set error for both learning algorithms can increase with training set size. However, we show in this paper that for uniform sampling and either algorithm, the expected error is a non-increasing function of training set size. For uniform sampling distributions, we also characterize the priors for which the expected error of the Bayes-optimal algorithm stays constant. In addition, we show that when the target function is fixed, expected off-training-set error can increase with training set size if and only if the expected error averaged over all targets decreases with training set size. Our results hold for arbitrary noise and arbitrary loss functions.
Download Info
To our knowledge, this item is not available for
download. To find whether it is available, there are three
options:
1. Check below under "Related research" whether another version of this item is available online.
2. Check on the provider's web page
whether it is in fact available.
3. Perform a search for a similarly titled item that would be
available.
Publisher Info
Paper provided by Santa Fe Institute in its series Working Papers with number
95-02-023.