0 Assigned ID: 1819 ----- Attempting Submission 1 (Fri Apr 09 04:29:05 GMT 2021) ----- HST Phase I Proposal 1819 successfully submitted. Receipt: # 1819-1 ----- Attempting Submission 2 (Fri Apr 09 18:42:09 GMT 2021) ----- HST Phase I Proposal 1819 successfully submitted. Receipt: # 1819-2 ----- Attempting Submission 3 (Fri Apr 09 20:20:18 GMT 2021) ----- In this proposal, we aim to train an artificial HST using Generative Adversarial Networks (GANs). Deep generative models are powerful unsupervised tools in learning true distributions of data sets. We will develop and publicly release an optimized deep learning framework using GANs which learns from the deepest HST archival galaxy images in single and multiple optical and NIR wavebands and produces HST-like observations given lower resolution ground-based images. We will quantify the confidence on the enhanced products given various combinations of input data with different qualities. This enables us to identify less represented classes of objects (if any) which could benefit from future HST observations. The enhanced images can be used by the community in various ways. Here, we mainly focus on the applications for weak lensing cosmology. With the images boosted in depth and/or resolution, we will be increasing the number of usable sources for weak lensing and will carefully quantify the gain in a) the deblending fraction, b) the photometry and c) photometric redshifts, as well as in d) the galaxy shape measurements. With these improvements, we can put tighter constraints on the cosmological parameters measurable through weak lensing. Our team includes experts in weak lensing cosmology, deep learning, and observers with extensive experience with HST data. In alphabetical order: - Dr. Behnam Darvish is an expert in galaxy evolution science and machine learning techniques. He has extensive experience analyzing both ground-based and HST data in the COSMOS and CANDELS fields. - Dr. Agnes Ferte is a theoretical cosmologist expert in weak lensing. She co-leads the Dark Energy Survey analysis team dedicated to test the different alternatives of the standard model of cosmology. She is also an expert member of the LSST and Roman weak lensing experiments. - Dr. Shoubaneh Hemmati is an observational astronomer and a deep learning expert and has broad experience in galaxy data analysis. As a member of the Roman team she has extensively worked on using deep HST galaxy surveys to improve the redshift measurements of the future weak lensing experiments of LSST, Euclid and Roman. - Dr. Eric Huff is an observational cosmologist and a guru in weak gravitational lensing. He is also an expert in statistical inferences and deep learning. He has extensively worked on understanding weak lensing systematics in DES. He is also an expert member of the Euclid and Roman weak lensing experiments. - Dr. Bahram Mobasher is an expert in galaxy evolution science and has extensive experience in HST survey designs, observations and large data-set analysis. He is also an expert in multi-band SED analysis and galaxy physical property measurements. - Dr. Jason Rhodes is an expert in weak lensing cosmology. He has a comprehensive knowledge of current and future cosmology missions, leads the US Euclid science team, and is the JPL Project Scientist for the Roman Space Telescope. - Dr. Harry Teplitz is an expert in galaxy formation and evolution science. He has led many successful programs with the HST. He has extensive experience in calibrating, simulating, and/or archiving data for large cosmology experiments including Euclid, Roman, and SPHEREx. Large Scale Structure of the Universe Galaxies 0 Regular false false false false false SUPPORTED