Demonstration Notebooks Overview

lrdbenchmark ships a Markdown-based notebook set (converted from the original Jupyter notebooks) that mirrors the tutorial series published in this documentation. Each notebook is self-contained, well annotated, and follows the same progression as the narrative material.

Notebook Structure

The demonstration notebooks follow a progressive learning structure:

1. Data Generation & Visualization - Understanding stochastic models

2. Estimation & Validation - Learning estimator categories and statistical validation

3. Custom Models & Estimators - Library extensibility and custom implementations

4. Comprehensive Benchmarking - Full benchmarking system with contamination testing

5. Leaderboard Generation - Performance rankings and comparative analysis

Tutorial Series

The instructional material previously hosted exclusively as notebooks is now published as a structured tutorial sequence. For the canonical, documentation-friendly narrative, follow the multi-part series:

• Data Generation and Visualisation — Generating synthetic and realistic LRD data, including contamination scenarios

• Estimation and Statistical Validation — Estimating the Hurst parameter with statistical validation and uncertainty quantification

• Custom Models and Estimators — Extending lrdbenchmark with custom data generators and estimators

• Comprehensive Estimator Benchmarking — Running comprehensive and contamination-aware benchmarks

• Leaderboard Generation — Building stratified leaderboards and publishing results

Notebook storage format

To keep the repository lightweight and diff-friendly, notebooks are distributed as Markdown files in notebooks/markdown/ together with exported figures and data artefacts. They can be opened directly in editors that understand MyST/Markdown notebooks, or converted back to .ipynb format with tools such as Jupytext.

Getting Started with Notebooks

Prerequisites

• Python 3.10–3.12

• Jupyter Notebook or JupyterLab

• lrdbenchmark installed (pip install lrdbenchmark)

Converting back to .ipynb

1. Clone the repository:

   git clone https://github.com/dave2k77/lrdbenchmark.git
   cd lrdbenchmark
   

2. Install the optional tooling:

   pip install -e .
   pip install jupytext jupyter matplotlib seaborn
   

3. Convert and launch:

   jupytext --to notebook notebooks/markdown/01_data_generation_and_visualisation.md
   jupyter notebook notebooks/markdown/
   

4. Open the converted notebooks in order for the complete learning experience.

Notebook Features

• Self-contained: Each notebook runs independently

• Progressive complexity: Build from simple to advanced concepts

• Practical focus: Real-world workflows, not just API documentation

• Reproducible: Fixed seeds, saved outputs

• Well-documented: Extensive markdown and comments

• Visual: Rich plots and tables throughout

• Educational: Explain why, not just how

Output Files

Each notebook generates the same artefacts as the original .ipynb versions:

• Visualisations: High-quality plots saved to outputs/.

• Data files: CSV/JSON exports of results.

• Performance metrics: Detailed analysis tables.

• Publication-ready figures: LaTeX/PNG formats.

Cross-References

The notebooks are designed to work together: - Notebook 1 → Notebook 2: Use generated data for estimation - Notebook 2 → Notebook 3: Apply estimators to custom models - Notebook 3 → Notebook 4: Include custom components in benchmarks - Notebook 4 → Notebook 5: Use benchmark results for leaderboards

This creates a complete workflow from data generation to performance analysis.

Support and Contributing

• Issues: Report problems with notebooks on GitHub Issues

• Discussions: Ask questions on GitHub Discussions

• Contributing: Submit improvements via pull requests

The notebooks are actively maintained and updated with each library release.