The weirdest SDSS galaxies: results from an outlier detection algorithm

Abstract

How can we discover objects we did not know existed within the large datasets that now abound in astronomy? We present an outlier detection algorithm that we developed, based on an unsupervised Random Forest. We test the algorithm on more than two million galaxy spectra from the Sloan Digital Sky Survey and examine the 400 galaxies with the highest outlier score. We find objects which have extreme emission line ratios and abnormally strong absorption lines, objects with unusual continua, including extremely reddened galaxies. We find galaxy-galaxy gravitational lenses, double-peaked emission line galaxies, and close galaxy pairs. We find galaxies with high ionisation lines, galaxies which host supernovae, and galaxies with unusual gas kinematics. Only a fraction of the outliers we find were reported by previous studies that used specific and tailored algorithms to find a single class of unusual objects. Our algorithm is general and detects all of these classes, and many more, regardless of what makes them peculiar. It can be executed on imaging, time-series, and other spectroscopic data, operates well with thousands of features, is not sensitive to missing values, and is easily parallelisable.

An automatic taxonomy of galaxy morphology using unsupervised machine learning

“…Unsupervised learning has already found application in astronomy, particularly in the estimation of photometric redshifts (Geach 2012;Way & Klose 2012;Carrasco Kind & Brunner 2014), object classification from photometry or spectroscopy (D'Abrusco et al 2012;in der Au et al 2012;Fustes et al 2013), finding galaxy clusters using catalogue data (Ascaso, Wittman & Benítez 2012) and searching for outliers in SDSS galaxy spectra (Baron & Poznanski 2016). Work by Schutter & Shamir (2015) presents computer vision techniques to identify galaxy types (see also Banerji et al 2010).…”

Evidence of ongoing AGN-driven feedback in a quiescent post-starburst E+A galaxy

“…However, there is no direct link between these outflows and the specific evolutionary phase of E+A galaxies. SDSS J132401.63+454620.6 was found as an outlier galaxy by the anomaly detection algorithm of Baron & Poznanski (2017). This galaxy is one of the 400 spectroscopically weirdest galaxies in the 12th data release (DR12) of the Sloan Digital Sky Survey (SDSS).…”

ConvoSource: Radio-Astronomical Source-Finding with Convolutional Neural Networks

“…Ackermann et al [14] investigated the use of deep convolutional neural networks (CNNs) and transfer learning in the automatic visual detection of galaxy mergers and found them to perform significantly better than current state-of-the-art merger detection methods. An outlier detection technique has also been developed using an unsupervised random forest algorithm and found to be successful in being able to detect unusual objects [15]. Gheller et al [16] developed COSMODEEP, a CNN to detect extended extragalactic radio sources in existing and upcoming surveys, which proved to be accurate and fast in detecting very faint sources in the simulated radio images.…”