ASGN

The official implementation of the ASGN model. Orginal paper: ASGN: An Active Semi-supervised Graph Neural Network for Molecular Property Prediction. KDD'2020 Accepted.

Project Structure

• base_model: Containing SchNet and training code for QM9 and OPV datasets.

• rd_learn: A baseline using random data selection.

• geo_learn: Geometric method of active learning like k_center.

• qbc_learn: Active learning by using query by committee.

• utils: Dataset preparation and utils functions.

• baselines: Active learning baselines from google's implementation.

• single_model_al: contains several baseline models and our method ASGN (in file wsl_al.py)

• exp: Experiments loggings.

How to learn

• You need to modify self.PATH in config.py depending on your environment.

1. qm9 download (below link)
https://figshare.com/articles/dataset/Data_for_6095_constitutional_isomers_of_C7H10O2/1057646?backTo=/collections/Quantum_chemistry_structures_and_properties_of_134_kilo_molecules/978904
2. PYTHONPATH=. python utils/pre/qm9_predata.py
3. PYTHONPATH=. python utils/pre/pre_qm.py
4. PYTHONPATH=. python single_model_al/wsl_al.py 

IDEAS

• Swav
  • Sinkhorn problem end2end 로 바꾸기.
• Multiple clustering
  • 다양한 기준으로 클러스터링한다?
• Signal from pseudo label
  • student 성능 기반으로 signal 얻기