Skip to content
Code for paper "Dimensionality-Driven Learning with Noisy Labels" - ICML 2018
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md
callback_util.py
complexity_plot.py
datasets.py
lass_tf.py
lid_plot.py
loss.py
loss_acc_plot.py
models.py
representation_plot.py
resnet.py
train_models.py
util.py

README.md

Code for ICML 2018 paper "Dimensionality-Driven Learning with Noisy Labels".

News: Issues fixed on CIFAR-10. 07/11/2018

1. Train DNN models using command line:

An example:

python train_model.py -d mnist -m d2l -e 50 -b 128 -r 40 

-d: dataset in ['mnist', 'svhn', 'cifar-10', 'cifar-100']
-m: model in ['ce', 'forward', 'backward', 'boot_hard', 'boot_soft', 'd2l']
-e: epoch, -b: batch size, -r: noise rate in [0, 100]

2. Run with pre-set parameters in main function of train_model.py:

    for dataset in ['mnist']:
        for noise_ratio in ['0', '20', '40', '60']:
            args = parser.parse_args(['-d', dataset, '-m', 'd2l',
                                      '-e', '50', '-b', '128',
                                      '-r', noise_ratio])
            main(args)

Requirements:

tensorflow, Keras, numpy, scipy, sklearn, matplotlib

You can’t perform that action at this time.