ErikEnglesson / GJS

The official code for the NeurIPS 2021 paper Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels (https://arxiv.org/abs/2105.04522)
21 stars 1 forks source link

Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels

The official code for the NeurIPS 2021 paper Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels

Environment Setup

Create conda environment, activate environment, and install additional pip packages

conda env create -f gjs_env.yml -n gjs
conda activate gjs
python -m pip install -r requirements.txt

Running Experiments

Please check scripts/ folder for yaml files corresponding to different experiments.

For example, to run JS on 40% symmetric noise on the full CIFAR-10 training set, run the following

python train.py -c scripts/C10/sym/js-40.yaml \
                --data_dir /path/to/dataset/

or GJS on 20% asymmetric noise on CIFAR-100

python train.py -c scripts/C100/asym/gjs-20.yaml \
                --data_dir /path/to/dataset/

or GJS on WebVision

python train.py -c scripts/WebVision/gjs.yaml \
                --data_dir /path/to/dataset/