embedding-dim

A Python library for estimating the embedding dimension of time series based on symbolic dynamics, entropy rate (via Lempel-Ziv complexity), and predictability (Pi_max).

Installation

git clone https://github.com/CoMuNeLab/embedding-dim.git
cd embedding-dim
pip install .

Usage

Option 1: via Python module

python -m scripts.run_dimension <input_csv> <output_prefix> <points> <tau_min> <tau_max> <max_m> [--scale SCALE]

Example:

python -m scripts.run_dimension data/lorenz_x.csv output 1000 1 3 10 --scale 1.0

Option 2: CLI alias (after installing with entry point)

embedding-dim <input_csv> <output_prefix> <points> <tau_min> <tau_max> <max_m> [--scale SCALE]

Example:

embedding-dim ../data/lorenz_x.csv output 1000 1 3 10

Arguments

Argument	Description
input_csv	Path to a CSV file with 1D time series
output_prefix	Prefix for output CSV and image files
points	Number of data points to use
tau_min	Minimum delay value τ
tau_max	Maximum delay value τ
max_m	Maximum embedding dimension
--scale	(Optional) Symbolization grid scale

Project Structure

├── data
│   └── lorenz_x.csv
├── pyproject.toml
├── README.md
├── scripts
│   └── run_dimension.py
└── src
    └── embedding_dim
        ├── dimension.py
        ├── entropy_rate.py
        ├── __init__.py
        ├── pi_max.py
        ├── substring.py
        └── symbolization.py

References

• V. d'Andrea et al., Symbolic dynamics for dimensionality estimation
• Tria et al., Predictability of human mobility, Nature (2012)