Revision history[back]
click to hide/show revision 1
Revision n. 1

Apr 28 '11 at 12:15

Brian%20Vandenberg's gravatar image

Brian Vandenberg
809213746

Conceptually, recurrent neural networks are not at all new. However, they've been largely relegated to academia because they were intractable to learn.

Some recent advances in hessian-free methods have made them more viable, though the authors' paper is awaiting publication.

click to hide/show revision 2
Revision n. 2

Apr 28 '11 at 12:16

Brian%20Vandenberg's gravatar image

Brian Vandenberg
809213746

Conceptually, recurrent neural networks are not at all new. However, they've been largely relegated to academia because they were intractable to learn.

Some recent advances in hessian-free methods have made them more viable, though the authors' paper is awaiting publication.

I list this because it's an emerging area that has a lot of potential if RNNs can be made to work well.

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.