Our take
- Activity
- 31
- Community
- 15
- Issues
- 21
- Pull requests
- 20
+1 stars in the last 30 days
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
311Forks
71Change since Nov 16, 2026: 0 stars·0 forks
Overview
nolds provides implementations of nonlinear measures used to analyze dynamical systems from one-dimensional time series. It helps researchers and data analysts quantify properties such as complexity, chaotic behavior, fractal dimension, and long-term memory without implementing the underlying methods themselves.
The library is also intended as a learning resource: its measure documentation explains algorithms and points to relevant references. It is a good fit when you need established measures for exploratory or research analysis, rather than a general-purpose time-series forecasting toolkit.
Key Features
- Estimates sample entropy to characterize time-series complexity.
- Calculates correlation dimension as a measure related to fractal structure and complexity.
- Implements Rosenstein and Eckmann approaches for estimating Lyapunov exponents.
- Provides rescaled-range and detrended fluctuation methods for Hurst-related analysis.
- Includes generalized Hurst exponent implementations for multifractal time series.
- Documents algorithms and references alongside the measure interfaces.
- Offers example scripts and unit tests for trying and checking the package.
Use Cases
- Researchers can compare complexity or dimensional properties across experimental time series.
- Dynamical-systems analysts can estimate Lyapunov exponents when investigating chaos and predictability.
- Quantitative analysts can examine long-term memory in sequential data, including financial time series.
- Students can use the documented implementations to learn how common nonlinear measures are calculated.
Getting Started
Install from PyPI:
pip install nolds
See the README and documentation for usage and measure details.
Considerations
- The core requirement is NumPy. Optional functionality needs scikit-learn for RANSAC fitting, quantumrandom for true random numbers, and matplotlib for plotting examples.
- The README warns that
mfhurst_dmis intended primarily for reproducing the referenced authors' results, and recommendsmfhurst_bfor general use. - The project is described as its author's side project, so responses to questions and issues may take days, weeks, or months.
- The README describes support for Python 2.7 and Python 3.4 or later. Check the repository documentation for current compatibility details before adopting it in a new environment.
Found this useful?
Share it with someone who would like nolds.