Skip to content

Latest commit

 

History

History
50 lines (32 loc) · 1.63 KB

README.md

File metadata and controls

50 lines (32 loc) · 1.63 KB
Raw

Use Perturbations when Learning from Explanations

Code accompanying submission titled "Use Perturbations when Learning from Explanations" to NeuriIPS 2023.

Algorithm implementations can be found in src/module and dataset implementations in src/datasets.
Hyperparameter searches and best hyperoparameters are documented in src/configs/expts.py.

How to run the code

Dependencies

  • Python and Conda

  • Setup the conda environment IBPEx by running:

    conda env create -f setup_dependencies.yml

    If you want to make use of your GPU, you might have to install a cuda-enabled pytorch version manually. Use the appropriate command provided here to achieve this. Before running code, activate the environment IBPEx.

    source activate IBPEx

Datasets

  • We have one synthetic dataset (decoy-MNIST) and two real dataset (ISIC and Plant phenotyping).
  • The default experiment settings for each dataset are in src/configs/dataset_defaults.py

Algorithms

  • We have 5 methods (RRR, IBP-Ex, PGD-Ex, CDEP, CoRM) in src/module, whose parent module is base_trainer.py.
  • The default experiment settings for each method are in src/configs/alg_defaults.py

Experiments

  • Run the code with the best hyperparameter setting we found for each dataset in src/configs/expts.py.

  • For example, to train the model using our IBP-Ex method on the Plant phenotyping dataset,

    python main.py --name 'ibp_plant_expt3'

  • View all possible command-line options by running

    python main.py --help