I would appreciate if you could devote some of your attention
to explaining the shape of the learning curves seen in statistical
Here below is a link to an extensive experimental study.
These learning curves are different than those I expect from
statistical learning theory
and it would be great to have some model explaining them.
I can think of no better crowd than Pascal2 to answer this question.
Learning to Translate: A Statistical and Computational Analysis
Marco Turchi, Tijl De Bie,Cyril Goutte, and Nello Cristianini
Advances in Artificial Intelligence
Volume 2012 (2012), Article ID 484580, 15 pages
The article can be found here: