To bring this full circle we can look at statistical inference as providing the theoretical basis for many aspects of algorithmic learning theory, from understanding and modeling the data to evaluating and improving the performance of learning algorithms.