scofield7419 / drga

Code for the paper Optimizing attention for sequence modeling via reinforcement learning.

DRGA： deep reinforcement learning guided attention

This repository includes the code of the paper Optimizing attention for sequence modeling via reinforcement learning published at IEEE TNNLS.

---

Requirement

python>=3.6
tensorflow
tflearn
numpy

Datasets

Text classifications:

• Movie Review (MR)
• AGnews
• Subjectivity (SUBJ)
• Stanford Sentiment Treebank (SST)

Word embedding

prepare the glove word embedding at emb file:

glove.6B.100d.txt

Running

python core/main.py

Citation

If you use this work, please kindly cite:

@article{FeiZRJ22,
  author       = {Hao Fei and
                  Yue Zhang and
                  Yafeng Ren and
                  Donghong Ji},
  title        = {Optimizing Attention for Sequence Modeling via Reinforcement Learning},
  journal      = {{IEEE} Trans. Neural Networks Learn. Syst.},
  volume       = {33},
  number       = {8},
  pages        = {3612--3621},
  year         = {2022},
  url          = {https://doi.org/10.1109/TNNLS.2021.3053633}
}

---

License

The code is released under Apache License 2.0 for Noncommercial use only.