diff options
author | Thomas Mesnard <thomas.mesnard@ens.fr> | 2015-12-20 10:58:07 +0100 |
---|---|---|
committer | Thomas Mesnard <thomas.mesnard@ens.fr> | 2015-12-20 10:58:07 +0100 |
commit | e4d1c9c632158d7f40d26567d10c558c8815b84d (patch) | |
tree | 0e823d931ea18cbd69fbad05c7cd17baeba0c528 /mohammad/README.md | |
parent | 2142359b779519dab8b31a7b04c17c8539942a77 (diff) | |
download | pgm-ctc-e4d1c9c632158d7f40d26567d10c558c8815b84d.tar.gz pgm-ctc-e4d1c9c632158d7f40d26567d10c558c8815b84d.zip |
Add some random code
Diffstat (limited to 'mohammad/README.md')
-rw-r--r-- | mohammad/README.md | 33 |
1 files changed, 33 insertions, 0 deletions
diff --git a/mohammad/README.md b/mohammad/README.md new file mode 100644 index 0000000..8c553c3 --- /dev/null +++ b/mohammad/README.md @@ -0,0 +1,33 @@ +CTC-Connectionist Temporal Classification +========================================= + +-CTC cost is implemented in pure [Theano](https://github.com/Theano/Theano). + +-Supports mini-batch. + +-Supports both normal- and log-scale. + +-"apple" problem (two same consecutive labels) is solved. + +-Test file is implemented using [Blocks](https://github.com/bartvm/blocks). + + + + +Reference +========= +Graves, Alex, et al. "Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks." Proceedings of the 23rd international conference on Machine learning. ACM, 2006. + + +Credits +======= +Theano implementation of CTC by [Shawn Tan](https://github.com/shawntan/rnn-experiment/) + +Theano implementation of CTC by [Rakesh Var](https://github.com/rakeshvar/rnn_ctc) + + +Special thanks to +================= +[Kyle Kastner](https://github.com/kastnerkyle) + +Pascal Lambdin |