This repository contains an implementation of Mitigating Label Noise through Data Ambiguation to be presented at AAAI-24. Please cite it as follows:
@misc{lienen2023mitigating,
title={Mitigating Label Noise through Data Ambiguation},
author={Julian Lienen and Eyke Hüllermeier},
year={2023},
eprint={2305.13764},
archivePrefix={arXiv},
primaryClass={cs.LG}
}
To install all required packages, you need to run
pip install -r requirements.txt
The code has been tested using Python 3.9 on Ubuntu 2*.* systems. We trained our models on machines with Nvidia GPUs (we tested CUDA 10.1, 11.1 and 11.6). Furthermore, we recommend to use Python virtual environments to get a clean Python environment for the execution without any dependency problems.
As a required prerequisite, the config.ini needs to be populated with parameters to set the output directory (BASE_PATH), a directory for temporary artifacts (TMP_PATH) and an output directory for plots (PLOT_DIR).
All datasets except for CIFAR-10(0)N, WebVision and Clothing1M are downloaded automatically. Webvision is available here, whereas access to Clothing1M has to be explicitly granted by the owner. CIFAR-10(0)is available here. All data needs to be stored in the specified --data_dir given as parameter to the training script (see next section).
For the training and evaluation, you have to call the following function (e.g., for CIFAR-10 with 25 % symmetric synthetic noise for our loss):
CUDA_VISIBLE_DEVICES=<the numeric ID(s) of your CUDA device(s)> python train.py --dataset=cifar10 --model resnet34 --seed 0 --loss RDA --adaptive_lrvar2 --adaptive_lrvar2_start_beta 0.75 --lrvar2_beta 0.6 --adaptive_lrvar2_type cosine --lr 0.02 --decay_type cosine --label_noise 0.25
--help allows for printing out all parameter options. All results presented in the paper were computed based on the training scripts train.py.
Our code uses the Apache 2.0 License, which we attached as LICENSE file in this repository.
Feel free to re-use our code. We would be happy to see our ideas put into practice.