PySDKit 0.4.56

Highlights

This release grows entropy of modes, adds underdetermined blind source separation, and ships the Read the Docs site that the gallery and API now live on.

• Entropy (pysdkit.entropy) — approximate, fuzzy, dispersion, spectral, distribution, increment, slope, and symbolic-dynamic entropy, plus Costa / composite / refined-composite multiscale variants of the measures that support them. Use them as scalar complexity features on a raw record or on each IMF after EMD / VMD.
• BSS (pysdkit.utils.bss) — Yu's frequency-energy clustering of the STFT (Shock and Vibration 2019; MATLAB pack originally named YGBSS). Underdetermined mixtures with delay are separated by binary frequency masks and inverse STFT; not EMD-style sifting.
• Documentation — Sphinx + pydata-sphinx-theme on pysdkit.readthedocs.io. Gallery scripts are Python (no notebooks in examples/), figures are cached for Read the Docs, and the navbar can switch latest / stable / 0.4.56.

Also on PyPI as 0.4.56: pip install -U PySDKit.

What's new

• Entropy (pysdkit.entropy)
  • approximate_entropy, fuzzy_entropy (+ multiscale_fuzzy_entropy)
  • dispersion_entropy (+ multiscale_dispersion_entropy)
  • spectral_entropy, distribution_entropy, increment_entropy, slope_entropy, symbolic_dynamic_entropy
  • composite_multiscale_sample_entropy, refined_composite_multiscale_sample_entropy
  • Existing permutation / sample entropy kept; shared coarse-graining helpers in _coarse_grain.py
  • Tests: pysdkit/tests/test_entropy.py
  • Gallery: examples/utils/entropy.py
• Blind source separation (pysdkit.utils.bss)
  • BSS / bss — frequency-energy clustering, cosine masks, inverse STFT
  • SDOF / MAC helpers for the modal-identification examples
  • Tests: pysdkit/tests/test_bss.py
  • Gallery: examples/utils/bss.py
• Docs site
  • User guide, API pages per family, About (cite / maintainers / license), Release notes from GitHub Releases
  • Sphinx-Gallery from examples/**/*.py; CONTRIBUTING.md for adding an algorithm
• Removed
  • pysdkit.models (KNN / PCA stubs). pysdkit.data.test_pca is unchanged.

Quick start

from pysdkit import VMD
from pysdkit.data import test_vmd
from pysdkit.entropy import spectral_entropy, sample_entropy
from pysdkit.utils import BSS

t, x, fs = test_vmd()
imfs = VMD(alpha=2000, K=4, tau=0.0, tol=1e-7).fit_transform(x)
print(spectral_entropy(imfs[0]), sample_entropy(imfs[0], m=2, r=0.2))

# Underdetermined BSS: mix, cluster frequency energy, recover sources
# y = BSS(num=2).fit_transform(mixture)  # see examples/utils/bss.py

Notes

• Import entropy as from pysdkit.entropy import … (not top-level pysdkit).
• Compare entropy values within one measure; SpecEn is scaled to [0, 1], IncrEn is in nats.
• BSS is a separator, not a decomposer: it returns recovered sources, not IMFs.
• Docs: latest tracks main; this tag is 0.4.56 once that version is built on Read the Docs.

Install

pip install -U PySDKit

Full Changelog: 0.4.55...0.4.56