Skip to content

API Reference

Feature pages

Page Contents
Meta Extractor, Bins
Variability Amplitude, BeyondNStd, Chi2Pvar, Eta, EtaE, ExcessVariance, Kurtosis, …
Linear trend LinearFit, LinearTrend
Time sampling Duration, ObservationCount, TimeMean, TimeStandardDeviation, …
Periodogram Periodogram
Non-linear parametric fits BazinFit, LinexpFit, VillarFit
Multiband parametric fit RainbowFit
Multiband ColorOfMaximum, ColorOfMedian, ColorOfMinimum, ColorSpread
Detection-based FluxNNotDetBeforeFd, MagnitudeNNotDetBeforeFd

All feature extractor classes share the same calling interface: __call__ for a single light curve and many for a batch.

Common attributes

Every extractor exposes two read-only attributes:

Attribute Type Description
names list[str] Output column names, one per extracted value
descriptions list[str] Human-readable description of each output 
import light_curve as licu

ext = licu.Extractor(licu.Amplitude(), licu.LinearFit())
print(ext.names)        # ['amplitude', 'linear_fit_slope', ...]
print(ext.descriptions) # ['Half amplitude of magnitude sample', ...]

Common interface

__call__ — single light curve

extractor(t, m, sigma=None, *, fill_value=None, sorted=None, check=True, cast=False)

Extract features and return them as a numpy array.

Parameter Type Description
t np.ndarray (float32 or float64) Time moments
m np.ndarray Magnitudes or fluxes
sigma np.ndarray, optional Photometric errors; assumed unity if None
fill_value float or None Value for invalid features; raises if None
sorted bool or None Whether t is sorted; checked if None
check bool Validate inputs for NaNs / infinities
cast bool Allow non-ndarray input and automatic dtype casting

Returns np.ndarray (float32 or float64) — one value per extracted feature.

many — batch processing

extractor.many(lcs, *, fill_value=None, sorted=None, check=True, cast=False, n_jobs=-1)

Parallel feature extraction over multiple light curves. Equivalent to:

np.stack([extractor(*lc, fill_value=fill_value, sorted=sorted, check=check)
          for lc in lcs])
Parameter Type Description
lcs list of (t, m) or (t, m, sigma) tuples, or Arrow array Input light curves
arrow_fields list[str or int] Required for Arrow input: field names/indices for [t, m] or [t, m, sigma]
fill_value float or None Same as __call__
sorted bool or None Same as __call__
check bool Same as __call__
n_jobs int Parallel workers; -1 uses all CPU cores

Returns np.ndarray of shape (N, n_features).

JSON serialization

Every extractor can be serialized to and from JSON, which preserves the class type and all constructor parameters.

to_json() — serialize

s = extractor.to_json()  # returns str

feature_from_json() — deserialize

import light_curve as licu

extractor = licu.feature_from_json(s)

Returns a JSONDeserializedFeature that behaves identically to the original extractor (same names, descriptions, and __call__ / many interface).

Round-trip example:

import light_curve as licu
import numpy as np

original = licu.Extractor(licu.Amplitude(), licu.LinearFit())
s = original.to_json()

restored = licu.feature_from_json(s)
assert restored.names == original.names

t = np.sort(np.random.default_rng(0).uniform(0, 100, 100))
m = np.random.default_rng(1).normal(15, 0.2, 100)
assert np.allclose(original(t, m), restored(t, m))