Since the advent of first generation modern astronomical observatories, the ensuing goal had been to obtain deeper and higher resolution observations of the sky. Though this can ultimately be attained…

This federal contract award was issued by a civilian agency, the Shared Services Center, to the Regents of the University of California at Riverside (UC Riverside), a prominent public research university and Hispanic-serving institution. The $189,449.00 grant is for research to use generative adversarial networks (GANs) to produce high-resolution images and spectra from lower-resolution multi-waveband data, with applications for future space telescopes like Euclid, SPHEREx, and Roman. The award does not have a set-aside designation, indicating open competition. UC Riverside has an extensive history of federal contract and grant awards, primarily focused on scientific research, technological innovation, and educational development across disciplines like environmental sciences, engineering, and computational sciences. The university's parent organization is the State of California Controllers Office, which has also been awarded federal contracts and grants, including for public health initiatives and advanced research through mechanisms like Other Transaction Agreements with the Defense Advanced Research Projects Agency (DARPA).

Since the advent of first generation modern astronomical observatories, the ensuing goal had been to obtain deeper and higher resolution observations of the sky. Though this can ultimately be attained with better instrumentation and larger observatories, these objectives would ultimately be limited by existing technology, most noticeably at longer wavelengths. The availability of deep high resolution data, specifically when available at multiple wavelengths in the form of broad-band and/or spectroscopic observations, has the power to determine the physical properties and evolutionary histories of galaxies. The field of machine learning has witnessed tremendous advances over the past decade with the development of the state-of-the-art hardware combined with neural network deep learning algorithms achieving super-human capabilities. This in particular extends to not only image classifications and feature identification but to quite interestingly machine generated data identical to the real observations, where its being readily utilized in various sectors. Generative adversarial networks (gans) are an unsupervised deep learning based generative approach that are now widely used in many diverse disciplines to create new data satisfying true distributions, remove noise, fill-in missing data or improve resolution among other applications. Distribution of galaxies in their multidimensional space of fluxes over the electromagnetic spectrum provides a good learning pool for astronomical gans. In this proposal we aim to use the NASA astrophysical database archive to train the distribution of galaxy images, seds and spectra over a wide wavelength range and use the trained models to produce high resolution images and spectra given lower resolution multi-waveband seds in the same wavelength range. IT has been shown that the broad-band colors have hidden information about higher resolution features in the spectra such as emission line properties. Gans provide the means to easily find these hidden correlations by classifying galaxies over the whole spectral energy distributions (seds). We will provide fine-tuned algorithms to increase the number of broad bands in the euclid and roman space telescope deep fields, increase the spatial resolution of the vera rubin observatory images, and increase the spectral resolution of spherex observations to match those of the keck. Benefits from these enhanced data products are countless, among which are less blending, better redshift measurements, accurate physical parameter estimates, more accurate weak lensing studies, and more precise cosmological measurements without the need for more telescope time.