qiuchili/qnn_torch

Quantum-inspired Neural Network

Focus on quantum-inspired representation learning and text classification.

Credits

This repository is supported by Benyou Wang, Qiuchi Li, Donghao Zhao, Chen Zhang and Amit.

Requirements

• PyTorch > 1.0.0
• Tensorflow
• NLTK

Implemented Models

• QDNN (Global Mixture)
• LocalMixtureNN (Local Mixture)
• MLLM
• SentiQDNN
• FastText
• CNN
• LSTM

Data

• Download the data at https://www.dropbox.com/s/zpu2wx5bq54agk8/data.zip?dl=0, and put the downloaded data folder in the root directory.
• Download the Glove embeddings at http://nlp.stanford.edu/data/glove.6B.zip, and put the downloaded embeddings folder in the root directory.

Usage

• Set up a configuration file and put it into directory /config. Details about how to configurate could be referred to in /config/config_qdnn.ini.
  • First, you should make it clear which network you will use, i.e. network_type.
  • Second, if you are using a network utilizing sentiment lexicon as external information such as SentiQDNN, we should set strategy in configuration file to multi-task.
  • Plus, you need specify other details like dataset_name, sentiment_dic_file, batch_size, lr, etc.
• Modify the configuration file you will use in run.py as

config_file = 'config/config_*.ini'

• Type in command line

python run.py 

Citations

If you find our work is useful, please kindly cite our papers:

@misc{li2018quantuminspired,
    title={Quantum-inspired Complex Word Embedding},
    author={Qiuchi Li and Sagar Uprety and Benyou Wang and Dawei Song},
    year={2018},
    eprint={1805.11351},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}
@misc{wang2019semantic,
    title={Semantic Hilbert Space for Text Representation Learning},
    author={Benyou Wang and Qiuchi Li and Massimo Melucci and Dawei Song},
    year={2019},
    eprint={1902.09802},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}