geoews 0.1.1

Information-geometric early warning signals: KL rate, Fisher–Rao geometry, geodesic acceleration.

geoews

Information-geometric early warning signals on sliding-window Gaussian models: KL divergence rate, Fisher–Rao step distances, and geodesic acceleration, plus classical benchmarks and optional data helpers.

Math matches the validated research implementation (sliding-window Gaussian fit, same KL and FR formulas, same regularization).

• Repository: github.com/vonixxxxx/geoews
• PyPI: pypi.org/project/geoews

Install

From PyPI (recommended):

pip install geoews

Upgrade:

pip install geoews --upgrade

Quick start — high-level API

ManifoldEWS fits sliding-window Gaussians, computes indicators, and runs a simple baseline-threshold detection:

import numpy as np
from geoews import ManifoldEWS

x = np.random.default_rng(0).standard_normal(500)
result = ManifoldEWS(window=40, cumul_window=30).fit(x).detect()
print(result)  # EWSResult with kl_rate, geodesic_acceleration, threshold, etc.

Lower-level API (same math, full control)

import numpy as np
from geoews.windows import estimate_gaussian_params, COVARIANCE_REGULARIZATION
from geoews.indicators import kl_rate, geodesic_acceleration

x = np.sin(np.linspace(0, 12, 2000))
times, mus, sigmas = estimate_gaussian_params(x, window_size=50, step=1)

kl = kl_rate(mus, sigmas)
acc = geodesic_acceleration(mus, sigmas, cumul_window=30)

kl_rate is the KL divergence rate between consecutive window Gaussians (alias of kl_divergence_rate).

Classical benchmarks

from geoews import variance_ews, acf_ews

times_v, var_series = variance_ews(x, window=50, step=1)
times_a, acf_series = acf_ews(x, window=50, step=1)

Optional datasets (local files)

load_ngrip and load_peter_lake expect paths to your own copies of the public datasets (Excel/CSV). See the examples/ notebooks in the repo for typical usage.

Constants

• COVARIANCE_REGULARIZATION defaults to 1e-6 (diagonal ridge on covariance / variance).

Development (from a git clone)

git clone https://github.com/vonixxxxx/geoews.git
cd geoews
pip install -e ".[dev]"
pytest

Citation

See CITATION.cff in the repository for citation metadata.

License

MIT — see LICENSE.