Research Project
Grant-in-Aid for Young Scientists (B)
We proposed a new deep learning methodology that enable fast and stable learning from limited amount training examples. We stacked convolutional layers using discriminative analytic solutions obtained by Fisher weight map to build multi-layer convolutional neural networks. Then we further fine tune the entire network by means of the standard backpropagation to quickly reach better local minima. Proposed method achieved state-of-the-art classification accuracy on some standard benchmarks, namely MNIST and STL-10.
All 2016 2015 2014
All Journal Article (3 results) (of which Peer Reviewed: 3 results, Acknowledgement Compliant: 3 results, Open Access: 1 results) Presentation (10 results) (of which Int'l Joint Research: 2 results, Invited: 3 results) Patent(Industrial Property Rights) (1 results)
IEICE Transactions on Information and Systems
Volume: E99.D Issue: 6 Pages: 1626-1634
10.1587/transinf.2015EDP7358
130005154735
Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)
Volume: 1 Pages: 585-590
Image and Video Technology, Lecture Notes in Computer Science (LNCS)
Volume: 9431 Pages: 682-694
10.1007/978-3-319-29451-3_54
