Variability class · AGN · sink

Other

The honest residual — extragalactic, exotic, and not-yet-classified.

Not every variable star in VSX is a star, and not every variable is cleanly typed. This bucket is the deliberate sink: active galactic nuclei, X-ray binaries, rare pulsators, and the genuinely unclassified.

35,004 active stars AGN · sink

Representative light curve

g · schematic

A representative light curve of red-noise flicker — the aperiodic wander of an accreting black hole. No period, no fold; that is the point.

What it is

Everything the eight-way rollup can't place into a physical family: active galactic nuclei and quasars (accreting supermassive black holes in distant galaxies), high- and low-mass X-ray binaries, unusual or under-studied pulsators, and stars marked simply variable or unclassified in VSX. The archive never deletes or hides them — VSX invalidations only change a status flag, and default queries filter to active stars.

Why keep a sink

A classifier that forces every object into a clean box is lying. Holding a principled "other" category is what lets the rest stay trustworthy, and it is exactly where the model's out-of-distribution and anomaly scores earn their keep. AGN are a good example of honest gaps elsewhere on the site: with no stellar distance and no single period, they appear only on the all-sky map in the explorer — never as a fabricated point on the colour-magnitude or period-amplitude diagrams.

The families within

Subtypes

Active galactic nucleus

AGN

Accreting supermassive black hole; stochastic, aperiodic optical flicker. Extragalactic.

Quasar

QSO

The luminous, distant end of the AGN family — variable on timescales of days to years.

X-ray binary

LMXB

A compact object accreting from a stellar companion; optical variability tracks the accretion flow.

Untyped variable

VAR

Confirmed variable, classification undetermined — a candidate for the classifier, not a verdict.

The varchive method

Finding the period

The same science code runs for every star. Here is how it behaves for this class — and where it can be fooled.

Mostly aperiodic, so period-finding is the wrong question. AGN variability is red noise — a damped random walk that produces seductive but spurious GLS peaks. varchive treats this class as the place where variability indices and the classifier's anomaly and out-of-distribution scores do the work, not the periodogram. When a real period does turn up (an X-ray binary's orbit, say), it is verified individually.

What to watch for

Red-noise wander mimics periodicity — a tall GLS peak on an AGN is almost always an artifact.
This is a heterogeneous bag by construction, so aggregate statistics over the whole class are close to meaningless.
It is a residual, not a physical family — a star landing here often just means VSX hasn't typed it yet.
Extragalactic objects have no meaningful stellar distance, so colour-magnitude and period-luminosity diagrams legitimately leave them out.

Keep exploring

Browse the Other population

Browse 35,004 stars → See them in the explorer