How predict-idlab/cct-cmc works
Key Components to Include:
Core Architecture Components
a) Data Layer:
• 'data' directory with two main datasets (IHDP and NLSM)
• Synthetic data generators from src/datasets
b) Model Implementation Layer:
• Three main components under 'src':
• conformal_metalearners (core implementation)
• cmc_metalearners (CMC specific implementation)
• wcp (Weighted Conformal Prediction)
c) Experimentation Layer:
• Notebooks directory containing experimental implementations
• Results directory storing output figures
Specific Implementation Details
• Model Components:
• T-Learner
• X-Learner
• S-Learner
• Base ML models (e.g., RandomForestRegressor)
• Processing Flows:
• Training flow
• Prediction flow
• Monte Carlo sampling process
• Conformal calibration process
Interaction Patterns
• Data flow between components
• Model training and prediction paths
• Evaluation feedback loops
Technical Details
a) Key Methods:
• fit()
• predict_int()
• predict_cps()
b) Configuration Options:
• adaptive_conformal
• pseudo_MC
• MC_samples