Justin Werfel, Xiaohui Xie, and H. Sebastian Seung. In, MIT Press, (2003)Discussion of learning curves for stochastic gradient descent.
Besides gradient based approaches, the paper shortly describes (with additional references) weight perturbation and node perturbation approaches..
Mark D. Smucker, James Allan, and Ben Carterette. CIKM '07: Proceedings of the sixteenth ACM conference on Conference on information and knowledge management, page 623--632. New York, NY, USA, ACM, (2007)