Skip to content

Quick Start

Run your first Foundation PLR experiment in 5 minutes.

Prerequisites

Ensure you have completed the installation steps.

Step 1: Activate Environment

cd foundation-PLR/foundation_PLR
source .venv/bin/activate

Step 2: Run a Classification Experiment

# Run with default configuration
python -m src.classification.flow_classification

This will:

  1. Load PLR data from the database
  2. Apply default preprocessing (outlier detection + imputation)
  3. Extract handcrafted features
  4. Train a CatBoost classifier
  5. Evaluate with bootstrap validation
  6. Log results to MLflow

Step 3: View Results

# Start MLflow UI
mlflow ui --port 5000

Open http://localhost:5000 to view experiment results.

MLflow UI showing experiment results

Configuration

Override defaults with Hydra:

# Change classifier
python -m src.classification.flow_classification classifier=XGBoost

# Change preprocessing
python -m src.classification.flow_classification \
    outlier_method=MOMENT-gt-finetune \
    imputation_method=SAITS

Available Methods

The pipeline supports:

Stage Methods Registry
Outlier Detection 11 methods configs/mlflow_registry/parameters/classification.yaml
Imputation 8 methods configs/mlflow_registry/parameters/classification.yaml
Classification 5 classifiers CatBoost recommended

Note: The registry is the single source of truth for method names.

Next Steps