Examples¶
Interactive Jupyter notebooks demonstrating the energy-repset framework, from basic usage to advanced multi-objective and constructive algorithm workflows.
Example 1: Getting Started¶
The simplest end-to-end workflow. Selects 4 representative months using a single objective (Wasserstein fidelity) and uniform weights.
Example 2: Feature Space Exploration¶
Chains statistical summaries with PCA dimensionality reduction, then uses multi-objective selection (Wasserstein + correlation + centroid balance) with KMedoids cluster-size weights.
Example 3: Hierarchical Seasonal Selection¶
Daily-resolution features with monthly-level selection under seasonal constraints. Uses GroupQuotaHierarchicalCombiGen to enforce one month per season.
Example 4: Comparing Representation Models¶
Same selection, three different representation models: Uniform, KMedoids cluster-size, and Blended (soft assignment). Compares how each distributes responsibility weights.
Example 5: Multi-Objective Exploration¶
Four-component objective with ParetoMaxMinStrategy vs WeightedSumPolicy. Includes Pareto front visualization and score contribution analysis.
Example 6: K-Medoids Clustering¶
Standalone k-medoids clustering demo. Selects representative months as cluster medoids with cluster-size-proportional weights, explores the effect of varying k.
Example 7: Constructive Algorithms¶
Three constructive algorithms -- Hull Clustering, CTPC, and Snippet -- that build solutions using their own internal objectives, bypassing the Generate-and-Test workflow.
Running the Notebooks¶
# Install the package in editable mode
pip install -e .
# Launch Jupyter
jupyter notebook docs/examples/