Four large-area Sunyaev-Zeldovich (SZ) experiments -- APEX-SZ, SPT, ACT, and Planck -- promise to detect clusters of galaxies through the distortion of Cosmic Microwave Background photons by hot (> 10^6 K) cluster gas (the SZ effect) over thousands o
f square degrees. A large observational follow-up effort to obtain redshifts for these SZ-detected clusters is under way. Given the large area covered by these surveys, most of the redshifts will be obtained via the photometric redshift (photo-z) technique. Here we demonstrate, in an application using ~3000 SDSS stripe 82 galaxies with r<20, how the addition of GALEX photometry (FUV, NUV) greatly improves the photometric redshifts of galaxies obtained with optical griz or ugriz photometry. In the case where large spectroscopic training sets are available, empirical neural-network-based techniques (e.g., ANNz) can yield a photo-z scatter of $sigma_z = 0.018 (1+z)$. If large spectroscopic training sets are not available, the addition of GALEX data makes possible the use simple maximum likelihood techniques, without resorting to Bayesian priors, and obtains $sigma_z=0.04(1+z)$, accuracy that approaches the accuracy obtained using spectroscopic training of neural networks on ugriz observations. This improvement is especially notable for blue galaxies. To achieve these results, we have developed a new set of high resolution spectral templates based on physical information about the star formation history of galaxies. We envision these templates to be useful for the next generation of photo-z applications. We make our spectral templates and new photo-z catalogs available to the community at http://www.ice.csic.es/personal/jimenez/PHOTOZ .
