Skip to content

Foundation PLR Documentation

Configuration

Configuration¶

Foundation PLR uses Hydra for configuration management.

Configuration Structure¶

configs/
├── defaults.yaml          # Main configuration
├── VISUALIZATION/         # Figure and plotting settings
├── mlflow_registry/       # MLflow metadata
└── ...

Key Configuration Values¶

Classification Parameters¶

CLS_EVALUATION:
  glaucoma_params:
    prevalence: 0.0354           # Disease prevalence (Tham 2014)
    tpAUC_sensitivity: 0.862     # Target sensitivity
    tpAUC_specificity: 0.821     # Target specificity

  BOOTSTRAP:
    n_iterations: 1000           # Bootstrap iterations
    alpha_CI: 0.95               # Confidence interval level

Visualization Settings¶

VISUALIZATION:
  dpi: 100
  figure_format: pdf

Overriding Configuration¶

Command Line¶

# Single override
python -m src.classification.flow_classification classifier=XGBoost

# Multiple overrides
python -m src.classification.flow_classification \
    classifier=CatBoost \
    CLS_EVALUATION.BOOTSTRAP.n_iterations=500

Configuration Files¶

Create a custom config file:

# configs/my_experiment.yaml
defaults:
  - defaults

classifier: CatBoost
outlier_method: MOMENT-gt-finetune
imputation_method: SAITS

Run with:

python -m src.classification.flow_classification --config-name=my_experiment

Environment Variables¶

Hydra supports environment variable interpolation:

data_path: ${oc.env:DATA_PATH,/default/path}

See Also¶