- Training and validation dataset preparing
- Explained model training
- Explained model evaluation
- Concept dataset preparing
- Concept model training (Concept discovered)
- Causal relationship calculation
- Causal relationship verification
conda create -n causal_concept python==3.6.13
conda activate causal_concept
pip install -r requirements.txt
- Dataset Generation for PVR In this stage, two PVR dataset is generated
python data\pvr_dataset_generation.py
- Explained Model Training In this stage, we will training 4 model on two dataset including PVR_MNIST and PVR_CIFAR10.
python model_training_for_pvr_task.py
- PVR Concept Dataset Generation In this stage, PVR Concept Datasets are generated.
python data\pvr_concepts_dataset_generation.py
- Concept Identification In this stage, concept is identified.
python concept_representation_identification.py
Thanks for the contributions from the below repos.
- CAV (https://github.com/rakhimovv/tcav)
- Concept based decision tree explanations (https://github.com/DataSystemsGroupUT/ACDTE/tree/master)
- ConceptSHAP (https://github.com/chihkuanyeh/concept_exp; https://github.com/arnav-gudibande/conceptSHAP)