Multi-parameter estimating photometric redshifts with artificial neural networks

We calculate photometric redshifts from the Sloan Digital Sky Survey Data Release 2 Galaxy Sample using artificial neural networks (ANNs). Different input patterns based on various parameters (e.g. magnitude, color index, flux information) are explored and their performances for redshift prediction are compared. For ANN technique, any parameter may be easily incorporated as input, but our results indicate that using dereddening magnitude produces photometric redshift accuracies often better than the Petrosian magnitude or model magnitude. Similarly, the model magnitude is also superior to Petrosian magnitude. In addition, ANNs also show better performance when the more effective parameters increase in the training set. Finally, the method is tested on a sample of 79, 346 galaxies from the SDSS DR2. When using 19 parameters based on the dereddening magnitude, the rms error in redshift estimation is sigma(z)=0.020184. The ANN is highly competitive tool when compared with traditional template-fitting methods where a large and representative training set is available.