/tensorflow-ASP-MTL

A Tensorflow implementation of Adversarial Shared-Private Model for Multi-Task Learning and Transfer Learning.

Primary LanguagePythonMIT LicenseMIT

tensorflow-ASP-MTL

A Tensorflow implementation of Adversarial Shared-Private Model. It's used in

  • Multi-Task Learning (ASP-MTL), see [1]
  • Transfer Learning (ASP-TL), see [2]

To train the model:

cd data/
unzip mtl-dataset.zip
cd ../
python main.py

Use Tensorboard to monitor the training or testing process:

tensorboard --logdir=summary/train

Reference

[1] Pengfei Liu, Xipeng Qiu and Xuanjing Huang, Adversarial Multi-task Learning for Text Classification

[2] Cen Chen, Yinfei Yang, Jun Zhou, Xiaolong Li and Forrest Sheng Bao, Cross-Domain Review Helpfulness Prediction based on Convolutional Neural Networks with Auxiliary Domain Discriminators

Acknowledgments

This project gets inspirations from the following projects: