Welcome to the lrdbenchmark documentation

lrdbenchmark delivers reproducible benchmarking, diagnostics, and reporting for long-range dependence (LRD) analysis. It combines stochastic data models, twenty estimators, contamination-aware preprocessing, stratified reporting, and provenance capture in one toolkit.

What lrdbenchmark provides

Twenty estimators with a unified API – 13 classical (temporal, spectral, wavelet, multifractal), 3 machine-learning, and 4 neural estimators.
Runtime profiles – switch between the lightweight quick profile (used automatically under pytest/CI) and the exhaustive full profile for publication-grade studies.
Robust benchmarking – contamination models, adaptive preprocessing, stratified summaries, non-parametric significance testing, and uncertainty calibration.
Nonstationarity testing – time-varying H generators (regime switching, continuous drift, structural breaks), critical regime models (OU, fractional Lévy, SOC), and structural break detection (CUSUM, Chow test, ICSS).
Surrogate data testing – IAAFT, phase randomization, and AR surrogates for hypothesis testing of LRD and nonlinearity.
Coverage probability analysis – Monte Carlo estimation of CI coverage with studentized bootstrap and calibration diagnostics.
Adaptive acceleration – automatic CPU mode with optional JAX → Numba → NumPy fallbacks and GPU support when requested.
Containerized experiments – Docker support for reproducible cloud/HPC benchmarking.
Documentation-first tutorials – the tutorial series now lives inside the docs and is mirrored by Markdown notebooks for interactive exploration.

Quick start

Install with pip install lrdbenchmark and see Quick Start Guide for end-to-end examples, including how to opt into the different runtime profiles.

Installation & Setup

API Reference

Theory & Background

For theoretical foundations and validation techniques, see the comprehensive examples in the API documentation and quickstart guide.

Examples & demos

Domain Guides

Domain-Specific Preprocessing

Demonstration materials

The original interactive curriculum is available in two forms:

Tutorial Series – the canonical documentation narrative.
Demonstration Notebooks Overview – guidance on using the Markdown notebook sources bundled in notebooks/markdown/.

Tutorial Series

Structured learning path through the lrdbenchmark workflow:

Tutorial Series