Skip to content

Latest commit

 

History

History
61 lines (35 loc) · 1.22 KB

README.md

File metadata and controls

61 lines (35 loc) · 1.22 KB

Representation Bottleneck

Overview

This repository is an implementation of the paper Discovering and Explaining the Representation Bottleneck of DNNs (arxiv), which was accepted as an Oral presentation at ICLR 2022.

Requirements

  • Python 3.9
  • pytorch 1.7.1
  • CUDA 11.0
  • numpy 1.19.5
  • torchvision 0.8.2

All models were trained on a single NVIDIA GeForce RTX 3090 GPU.

Usage

Training

Run the following shell script to train the models:

./train.sh

You can change the gpu by changing the --gpu_id argument in the script.

The models are saved in the checkpoints directory by default.

Compute interaction

Run the following shell script to compute interaction for the models:

./interaction.sh

You can uncomment the setting you want to run on top of the script.

The results are saved in the results directory by default.

Citation

If you use this project in your research, please cite it.

@article{deng2021discovering,
  title={Discovering and Explaining the Representation Bottleneck of DNNs},
  author={Deng, Huiqi and Ren, Qihan and Zhang, Hao and Zhang, Quanshi},
  journal={arXiv preprint arXiv:2111.06236},
  year={2021}
}