Skip to content

kylehkhsu/role-of-data

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

72 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Code accompanying On the role of data in PAC-Bayes bounds, AISTATS 2021

Installation

git clone https://github.com/kylehkhsu/role-of-data
cd role-of-data
conda env create -f environment.yml
conda activate role-of-data

Login to Weights & Biases

  1. Navigate to https://wandb.ai/authorize, login, and copy the key (KEY)
  2. wandb login KEY

Overview

data_dependent_prior_sgd.py runs the SGD-characterization experiments whose results are plotted in Figure 4.

data_dependent_prior_direct.py runs the direct bound optimization experiments whose results are plotted in Figure 5.

data_dependent_prior_ghost_sgd.py runs the experiments whose results are plotted in the right column of Figure 7 (Appendix).

data_dependent_prior_half_sgd.py runs the experiments whose results are plotted in Figure 8 (Appendix).

sweeps/ contains config files including all hyperparameters used for the presented results.

Running the code

You have two options:

  1. Simply run a script, e.g. python scripts/data_dependent_prior_sgd.py
  2. Use a config file in sweeps/ to run a wandb sweep (see this example). As-is, each file specifies a sweep over 50 seeds.

Citing this work

If you find the code and/or paper helpful for your research, please cite

@inproceedings{dziugaite2021role,
    title={On the role of data in PAC-Bayes bounds},
    author={Dziugaite, Gintare Karolina and Hsu, Kyle and Gharbieh, Waseem and Arpino, Gabriel and Roy, Daniel M},
    booktitle={International Conference on Artificial Intelligence and Statistics (AISTATS)},
    year={2021}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages