/banditnet

PyTorch implementation of BanditNet

Primary LanguagePython

PyTorch implementation of BanditNet

Dependencies

  • torch >= 0.4.0
  • tqdm
  • numpy
  • tensorboardX

Instructions to run

  1. mkdir data/; cd data
  2. Download CIFAR-10 (Link) in data/
  3. tar -xvf cifar-10-python.tar.gz
  4. cd ..; python preprocess_cifar.py
  5. cd code; python main.py

General Information

  • Track all progress using tensorboard:
    • tensorboard --logdir code/tensorboard_stuff --port 16006
    • http://localhost:16006
  • preprocess_cifar.py makes bandit_dataset from the original cifar dataset
    • Edit the file to update hyper_parameters like num_sample
  • All the hyper-parameters for the banditnet implementation can be set in the file code/hyper_params.py

References

Paper: https://www.cs.cornell.edu/people/tj/publications/joachims_etal_18a.pdf

BibTeX:

@InProceedings{Joachims/etal/18a,
  author = 	 {T. Joachims and A. Swaminathan and M. de Rijke},
  title = 	 {Deep Learning with Logged Bandit Feedback},
  booktitle =    {International Conference on Learning Representations (ICLR)},
  year = 	 2018}