Abstract
Probabilistic convergence results of online gradient descent algorithm have been obtained by many authors for the training of recurrent neural networks with innitely many training samples. This paper proves deterministic convergence of o2ine gradient descent algorithm for a recurrent neural network with nite number of training samples. Our results can be hopefully extended to more complicated recurrent neural networks, and serve as a complementary result to the existing probability convergence results. © Springer-Verlag Berlin Heidelberg 2007.
Cite
CITATION STYLE
Xu, D., Li, Z., Wu, W., Ding, X., & Qu, D. (2007). Convergence of gradient descent algorithm for a recurrent neuron. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4493 LNCS, pp. 117–122). Springer Verlag. https://doi.org/10.1007/978-3-540-72395-0_16
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
