Aims: We present an analysis of the effects of integrating galaxy morphological information in photometric redshift (photo-z) estimation with a custom support vector machine (SVM) classification package. We also present a comparison with other methods. Statistical correlations between galaxy shape information and redshift that are not degenerate with photometric band magnitudes would be evident through an improvement in the accuracy of photo-z estimations, or possibly even in a lack of significant loss of accuracy in light of the noise introduced by including additional parameters. Methods: SVM algorithms, a type of machine learning, utilize statistical learning theory and optimization theory to construct predictive models based on the information content of data in a way that can treat different input types symmetrically, which can be a useful estimator of the additional information contained in parameters, such as those describing the morphology of the galaxies. The custom SVM classification code we have developed is designated SPIDERz and is made available to the community. As test data we use imaging and five band photometric magnitudes from the All-wavelength Extended Groth Strip International Survey. Results: We find that for the data used this SVM algorithm results in a significantly decreased number of outliers and RMS error compared to some other considered techniques, and that the inclusion of morphological information does not have a statistically significant benefit for photo-z estimation with the techniques employed here, which is roughly in agreement with a previous analysis considering an artificial neural network method. We conclude that it is therefore likely a generic result for empirical redshift estimation techniques that the inclusion of morphological information does not improve metrics such as the RMS error or number of outliers.
↧