Nick Ball - Research - Galaxy Types from Artificial Neural Networks

The results on this page are described in more detail in Ball et al. (2004), also at astro-ph/0306390 . The catalog containing the assigned types is here.

The results involve the ability of the artifical neural network to predict types for galaxies. In particular the accurate prediction of morphological and spectral type using purely photometric parameters is possible. The broad variables are the galaxy type (or in principle any parameter) to be predicted, the parameter set used, and the architecture of the network. Each of these aspects is discussed below. There is also a further page on the results from the neural nets, with more details and plots than the paper here, as I have not found any other pages showing anything similar.

Dataset

The training sets are from the SDSS Data Release One (DR1), matched to the galaxies for which the types have been assigned. There is also a paper describing DR1, which extends the paper describing the Early Data Release.

Galaxy parameters

The following parameters, available from DR1 have been used. Further details are in the paper or in the SDSS SkyServer Schema browser.

1        Petrosian radius in r band
2        50 percent light radius in r (r_50)
3        90 percent light radius in r (r_90)
4        de Vaucouleurs profile radius in r
5        Exponential profile radius in r
6        de Vaucouleurs profile axial ratio in r
7        Exponential profile axial ratio in r
8        log likelihood of de Vaucouleurs galaxy light profile
9        log likelihood of exponential profile
10       galaxy surface brightness
11       concentration index r_50/r_90 in r
12-15    model u-g, g-r, r-i, i-z colours
16-19    Petrosian u-g, g-r, r-i, i-z colours
20-24    model u g r i z magnitudes
25-29    Petrosian u g r i z magnitudes

Morphological Type

Similarly, the eyeball morphological type can be predicted with a correlation of 0.93 and rms of 0.55. The range is 0 to 6 (elliptical to spiral).

0306390_figure1_colour.gif

Network type versus morphological type. Note that the the targets, assigned by human experts are to the nearest 0.5 in type.


An automated measure of morphological type being developed in the SDSS, known as TP[auto], which requires galaxy images of 25+ pixels in size, can also be predicted with 0.90 correlation.

Spectral Type

The networks are able to predict the eClass spectral type with a correlation of 0.95 and rms deviation of network type from target type of 0.06. The range of targets is approximately -1 (spiral galaxies) to 0.5 (elliptical). The sign of the eClass may be changed in future but this should have no effect on the results.

0306390_figure2_colour.gif

Network type versus eClass


This is using the parameter set 1 to 29. No subset of parameters performs as well, although some are close, e.g. the four model colours give 0.94.

Redshift

In principle any parameter describing the galaxies could be predicted if it is available in the training set, for examples redshift. Other groups have looked at this (e.g. Firth et al 2003, Tagliaferri et al 2002) and achieve comparable results.

0306390_figure3_colour.gif

Network redshift versus redshift. Some large scale structure is seen as vertical banding in the target redshifts.

General Neural Network Issues

There are numerous issues with the use of neural nets to perform these sorts of predictions - the nature of the neural nets is that you could be doing something wrong and they will still give reasonable, but not quite so good, results. Much of the DPhil work so far has been making sure this is avoided by exploring the issues so that confidence is gained in the predicted galaxy types.

Further aspects of the neural nets, including many plots and details not mentioned in the paper are here.


Last modified: Thu Jun 1 17:09:43 CDT 2006

Research
Home