Presented @ Graph Signal Processing Workshop 2023 by Nicolas Farrugia

Overview of the framework. Each image is augmented twice to produce two views. The edges represent probability transitions of pair of views. Thick edges represent high probability transitions (positive pairs) that are maximized toward 1, while dotted lines correspond to low probability transitions (negative pairs), minimized toward 0.

To run CLRW :
python3 main.py --epochs 100 --epochs2 100 --lr 0.1 --lr2 0.1 --tau 0.4 --datadir path_for_storing_data

To run SimCLR, add the argument --simclr

To use AutoAugment, add the argument --autoaugment