Code for 'FREDIS: A Fusion Framework of Refinemnet and Disambiguation for Unreliable Partial Label Learning' (ICML'23)
Requirements: Python 3.6.9, numpy 1.19.5, torch 1.9.1, torchvision 0.10.1.
or
conda env create -f environment.yaml
You need to:
- Download the datasets into './data/'.
- Run the following example:
python main.py --dataset cifar10 --noisy_rate 0.3 --partial_rate 0.3