The code for the ACL 2017 paper “Scalable Bayesian Learning of Recurrent Neural Networks for Language Modeling”
- Most of the experiments are implemented with Theano.
- The language modeling experiment on PTB is implemented with both Theano and Torch.
-
The data used in our paper can be downloaded here.
-
Running the python files can reproduce the results in the paper.
-
For the PTB dataset with successive minibatches, we extend wojzaremba's lua code. For the PTB dataset with random minibatches, we use the provided theano code to run the experiments.
Please cite our ACL paper in your publications if it helps your research:
@inproceedings{BayesianRNN_ACL2017,
title={Scalable Bayesian Learning of Recurrent Neural Networks for Language Modeling},
author = {Gan, Zhe and Li, Chunyuan and Chen, Changyou and Pu, Yunchen and Su, Qinliang and Carin, Lawrence},
booktitle={ACL},
Year = {2017}
}