GALAXY MORPHOLOGY WITHOUT CLASSIFICATION - SELF-ORGANIZING MAPS

We examine a general framework for visualizing data sets of high (grea ter than 2) dimensionality and demonstrate the framework by taking the morphology of galaxies at moderate redshifts as an example. The distr ibutions of various populations of such galaxies are examined in a spa ce spanned by four purely morphological parameters. Galaxy images are taken from the Hubble Space Telescope Wide Field Planetary Camera 2 in the I band (using the F814W filter). Since we have little prior knowl edge on how galaxies are distributed in morphology space, we use an un supervised learning method (a variant of Kohonen's self-organizing map s, or SOMs). This method allows the data to organize themselves onto a two-dimensional space while conserving most of the topology of the or iginal space. It thus enables us to visualize the distribution of gala xies and study it more easily. The process is fully automated, does no t rely on any kind of ''eyeball'' classification and is readily applic able to large numbers of images. We apply it to a sample of 2934 galax ies and find that the morphology correlates well with the apparent mag nitude distribution and, to a lesser extent, with color and bulge domi nance. The resulting map traces a morphological sequence similar to th e Hubble sequence, albeit two-dimensional. We use the SOM as a diagnos tic tool and rediscover a population of bulge-dominated galaxies with morphologies characteristic of peculiar galaxies. This result is achie ved without recourse to classification by eye. We also examine the eff ect of noise on the resulting SOM, and conclude that our results are r eliable down to an I magnitude of 24. We propose using this method as a framework into which more physical data can be incorporated as they become available. We hope that this method will lead to a deeper under standing of galaxy evolution.