Integrations

polars

Every function that accepts a pandas Series also accepts a polars Series. Rolling and delta results come back as a polars Series with the same name:

import polars as pl
from entroscope import shannon, permutation

s = pl.Series("interest", [10, 20, 15, 80, 90, 85, 88, 92] * 10)

shannon.compute(s)                  # float
permutation.rolling(s, window=20)   # polars Series named "interest"

Install with pip install "entroscope[polars]" (or just have polars installed).

scikit-learn

EntropyFeatures turns time-series windows into entropy features, so entropy drops straight into a scikit-learn pipeline. Each row of X is one window; each output column is one measure.

import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.pipeline import make_pipeline
from entroscope.features import EntropyFeatures

# Cut a long series into overlapping windows of 64 samples, 16 apart.
windows = np.lib.stride_tricks.sliding_window_view(signal, 64)[::16]

model = make_pipeline(
    EntropyFeatures(measures=("spectral", "permutation", "sample")),
    RandomForestClassifier(),
)
model.fit(windows, labels)

Pass per-measure settings with params:

EntropyFeatures(
    measures=("sample", "permutation"),
    params={"sample": {"m": 2, "r": 0.15}, "permutation": {"order": 4}},
)

For named DataFrame output, use scikit-learn's output API:

EntropyFeatures().set_output(transform="pandas").fit_transform(windows)
# columns: shannon_entropy, permutation_entropy, spectral_entropy, ...

Available measures: shannon, permutation, spectral, sample, approximate, differential. The transformer is stateless, so fit only validates the input. Install with pip install "entroscope[sklearn]".