Using transfer learning to detect galaxy mergers

We investigate the use of deep convolutional neural networks (deep CNNs) for automatic visual detection of galaxy mergers. Moreover, we investigate the use of transfer learning in conjunction with CNNs, by retraining networks first trained on pictures of everyday objects. We test the hypothesis that transfer learning is useful for improving classification performance for small training sets. This would make transfer learning useful for finding rare objects in astronomical imaging datasets. We find that these deep learning methods perform significantly better than current state-of-the-art merger detection methods based on nonparametric systems like CAS and GM$_{20}$. Our method is end-to-end and robust to image noise and distortions; it can be applied directly without image preprocessing. We also find that transfer learning can act as a regulariser in some cases, leading to better overall classification accuracy ($p = 0.02$). Transfer learning on our full training set leads to a lowered error rate from 0.038 $\pm$ 1 down to 0.032 $\pm$ 1, a relative improvement of 15%. Finally, we perform a basic sanity-check by creating a merger sample with our method, and comparing with an already existing, manually created merger catalogue in terms of colour-mass distribution and stellar mass function.

[1]  George Cybenko,et al.  Approximation by superpositions of a sigmoidal function , 1992, Math. Control. Signals Syst..

[2]  A. Szalay,et al.  Galaxy Luminosity Functions to z~1 from DEEP2 and COMBO-17: Implications for Red Galaxy Formation , 2005, astro-ph/0506044.

[3]  E. Suchman,et al.  The American soldier: Adjustment during army life. (Studies in social psychology in World War II), Vol. 1 , 1949 .

[4]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[5]  C. Lintott,et al.  Galaxy Zoo 1: data release of morphological classifications for nearly 900 000 galaxies , 2010, 1007.3265.

[6]  Ronald J. Buta,et al.  The Catalog of Southern Ringed Galaxies , 1994 .

[7]  E. Bell,et al.  The Optical and Near-Infrared Properties of Galaxies. I. Luminosity and Stellar Mass Functions , 2003, astro-ph/0302543.

[8]  G. Neugebauer,et al.  Ultraluminous infrared galaxies and the origin of quasars , 1988 .

[9]  H. D. Brunk,et al.  Statistical inference under order restrictions : the theory and application of isotonic regression , 1973 .

[10]  C. Lintott,et al.  Galaxy Zoo: morphologies derived from visual inspection of galaxies from the Sloan Digital Sky Survey , 2008, 0804.4483.

[11]  L. Hernquist,et al.  Gasdynamics and starbursts in major mergers , 1995, astro-ph/9512099.

[12]  Carnegie-Mellon,et al.  A Unified, Merger-driven Model of the Origin of Starbursts, Quasars, the Cosmic X-Ray Background, Supermassive Black Holes, and Galaxy Spheroids , 2005, astro-ph/0506398.

[13]  François Chollet,et al.  Xception: Deep Learning with Depthwise Separable Convolutions , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[14]  E. Treister,et al.  MAJOR GALAXY MERGERS ONLY TRIGGER THE MOST LUMINOUS ACTIVE GALACTIC NUCLEI , 2012, 1209.5393.

[15]  Christopher J. Conselice,et al.  The Relationship between Stellar Light Distributions of Galaxies and Their Formation Histories , 2003 .

[16]  Stuart Lynn,et al.  The Galaxy Zoo survey for giant AGN-ionized clouds: past and present black hole accretion events , 2011, 1110.6921.

[17]  Richard S. Ellis,et al.  Analysis of a complete galaxy redshift survey – II. The field-galaxy luminosity function , 1988 .

[18]  A. S. Szalay,et al.  Galaxy Zoo: the fraction of merging galaxies in the SDSS and their morphologies , 2009, 0903.4937.

[19]  F. Attneave,et al.  The Organization of Behavior: A Neuropsychological Theory , 1949 .

[20]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[21]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[22]  T. D. Matteo,et al.  Modelling feedback from stars and black holes in galaxy mergers , 2004, astro-ph/0411108.

[23]  Ben Hoyle,et al.  Measuring photometric redshifts using galaxy images and Deep Neural Networks , 2015, Astron. Comput..

[24]  Tom Fawcett,et al.  An introduction to ROC analysis , 2006, Pattern Recognit. Lett..

[25]  Antonio Torralba,et al.  Comparison of deep neural networks to spatio-temporal cortical dynamics of human visual object recognition reveals hierarchical correspondence , 2016, Scientific Reports.

[26]  Sander Dieleman,et al.  Rotation-invariant convolutional neural networks for galaxy morphology prediction , 2015, ArXiv.

[27]  Puragra Guhathakurta,et al.  The DEEP2 Galaxy Redshift Survey: Evolution of Close Galaxy Pairs and Major-Merger Rates up to z ~ 1.2 , 2004, astro-ph/0411104.

[28]  A. Toomre,et al.  Galactic Bridges and Tails , 1972 .

[29]  M. Geller,et al.  Minor Galaxy Interactions: Star Formation Rates and Galaxy Properties , 2007, astro-ph/0703729.

[30]  Sugata Kaviraj,et al.  Galaxy Zoo: Major Galaxy Mergers Are Not a Significant Quenching Pathway , 2017, 1708.00866.

[31]  Joel R. Primack,et al.  The effect of mass ratio on the morphology and time-scales of disc galaxy mergers: Effect of mass ratio on merger morphology , 2009, 0912.1590.

[32]  Welch Bl THE GENERALIZATION OF ‘STUDENT'S’ PROBLEM WHEN SEVERAL DIFFERENT POPULATION VARLANCES ARE INVOLVED , 1947 .

[33]  P. Madau,et al.  A NEW NONPARAMETRIC APPROACH TO GALAXY MORPHOLOGICAL CLASSIFICATION , 2003, astro-ph/0311352.

[34]  Christian Wolf,et al.  A new automatic method to identify galaxy mergers – I. Description and application to the Space Telescope A901/902 Galaxy Evolution Survey★ , 2011, 1109.6828.

[35]  Alessandro Caccianiga,et al.  The merger fraction of active and inactive galaxies in the local Universe through an improved non-parametric classification , 2013, 1303.0036.

[36]  Kurt Hornik,et al.  Approximation capabilities of multilayer feedforward networks , 1991, Neural Networks.

[37]  F. James Statistical Methods in Experimental Physics , 1973 .

[38]  R. Nichol,et al.  Quantifying the Bimodal Color-Magnitude Distribution of Galaxies , 2003, astro-ph/0309710.

[39]  David Schiminovich,et al.  The Star Formation and Extinction Coevolution of UV-Selected Galaxies over 0.05 < z < 1.2 , 2007, 0709.0730.

[40]  Oriol Vinyals,et al.  Qualitatively characterizing neural network optimization problems , 2014, ICLR.

[41]  Kunihiko Fukushima,et al.  Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position , 1980, Biological Cybernetics.

[42]  O. I. Wong,et al.  The green valley is a red herring: Galaxy Zoo reveals two evolutionary pathways towards quenching of star formation in early-and late-type galaxies , 2014, 1402.4814.

[43]  Mark Lacy,et al.  SHOCKED POSTSTARBUST GALAXY SURVEY. I. CANDIDATE POST-STARBUST GALAXIES WITH EMISSION LINE RATIOS CONSISTENT WITH SHOCKS , 2016, 1601.05085.

[44]  Anna K. Weigel,et al.  Stellar mass functions: methods, systematics and results for the local Universe , 2016, 1604.00008.

[45]  J. P. Jones,et al.  An evaluation of the two-dimensional Gabor filter model of simple receptive fields in cat striate cortex. , 1987, Journal of neurophysiology.

[46]  P. Natarajan,et al.  Major Galaxy Mergers and the Growth of Supermassive Black Holes in Quasars , 2010, Science.

[47]  S Marcelja,et al.  Mathematical description of the responses of simple cortical cells. , 1980, Journal of the Optical Society of America.

[48]  B. Garilli,et al.  zCOSMOS – 10k-bright spectroscopic sample - The bimodality in the galaxy stellar mass function: exploring its evolution with redshift , 2009, 0907.5416.

[49]  J. Daugman Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two-dimensional visual cortical filters. , 1985, Journal of the Optical Society of America. A, Optics and image science.

[50]  Daniel H. McIntosh,et al.  A First Estimate of the Baryonic Mass Function of Galaxies , 2003, astro-ph/0301616.

[51]  Joel R. Primack,et al.  Galaxy merger morphologies and time-scales from simulations of equal-mass gas-rich disc mergers , 2008, 0805.1246.

[52]  Allan Sandage,et al.  The velocity field of bright nearby galaxies. I - The variation of mean absolute magnitude with redshift for galaxies in a magnitude-limited sample , 1979 .

[53]  Maarten Schmidt,et al.  Space Distribution and Luminosity Functions of Quasi-Stellar Radio Sources , 1968 .

[54]  M. Whitlock Combining probability from independent tests: the weighted Z‐method is superior to Fisher's approach , 2005, Journal of evolutionary biology.

[55]  Nilotpal Chakravarti,et al.  Isotonic Median Regression: A Linear Programming Approach , 1989, Math. Oper. Res..

[56]  Edwin Simpson,et al.  Space Warps – I. Crowdsourcing the discovery of gravitational lenses , 2015, 1504.06148.