Awesome

Sequential GCN for Active Learning

Please cite if using the code: Link to paper.

Requirements:

python 3.6+

torch 1.0+

pip libraries: tqdm, sklearn, scipy, math

Run:

For running UncertainGCN on CIFAR-10 over 5 sampling stages of 1000 images:

python main.py -m UncertainGCN -d cifar10 -c 5 # Other available datasets cifar100, fashionmnist, svhn

CoreGCN, the geometric method that uses GCN training, can be run as:

python main.py -m CoreGCN -d cifar10 -c 5 # Other AL methods: Random, VAAL, CoreSet, lloss

Please have a look over the config file before running. Also, check the args of the code. CUDA-GPU implementation, not tested on CPU. Different random seed might produce different results.

Active Learning methods implemented:

Active Learning for Convolutional Neural Networks: A Core-Set Approach: https://arxiv.org/pdf/1708.00489.pdf [CoreSet]

Learning Loss for Active Learning: https://arxiv.org/pdf/1905.03677.pdf [lloss]

Variational Adversial Active Learning: https://arxiv.org/pdf/1904.00370.pdf [VAAL]

Contact

If there are any questions or concerns feel free to send a message at: r.caramalau18@imperial.ac.uk