GithubHelp home page GithubHelp logo

henghengxiong / rocl Goto Github PK

View Code? Open in Web Editor NEW

This project forked from kim-minseon/rocl

0.0 0.0 0.0 88.63 MB

Code for the paper "Adversarial Self-supervised Contrastive Learning" (NeurIPS 2020)

Home Page: https://sites.google.com/view/rocl2020

Shell 3.77% Python 96.23%

rocl's Introduction

RoCL-Adversarial self-supervised contrastive learning

This repository is the official PyTorch implementation of "Adversarial self supervised contrastive learning" by Minseon Kim, Jihoon Tack and Sung Ju Hwang.

Requirements

Currently, requires following packages

Training

To train the model(s) in the paper, run this command:

mkdir Data folder inside the RoCL

mkdir ./Data
python -m torch.distributed.launch --nproc_per_node=2 rocl_train.py --ngpu 2 --batch-size=256 --model='ResNet18' --k=7 --loss_type='sim' --advtrain_type='Rep' --attack_type='linf' --name=<name-of-the-file> --regularize_to='other' --attack_to='other' --train_type='contrastive' --dataset='cifar-10'

Evaluation

To evaluate my model linear evaluation and robustness, run:

./total_process.sh test <checkpoint-load> <name> <model type='ResNet18' or 'ResNet50'> <learning rate=0.1> <dataset='cifar-10' or 'cifar-100'>

Results

Our model achieves the following performance on :

Classification and robustness on CIFAR 10

Model name Accuracy robustness
RoCL ResNet18 83.71 % 40.27%

Citation

@inproceedings{kim2020adversarial,
  title={Adversarial Self-Supervised Contrastive Learning},
  author={Minseon Kim and Jihoon Tack and Sung Ju Hwang},
  booktitle = {Advances in Neural Information Processing Systems},
  year={2020}
}

rocl's People

Contributors

kim-minseon avatar jihoontack avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.