A Revolution for Astrophysical Turbulence with Machine Learning Methods

TURBULENCE IS OF PARAMOUNT IMPORTANCE IN DEVELOPING AN INFORMED UNDERSTANDING OF A MYRIAD OF PHYSICAL PHENOMENA. IT IS FOUND IN THE FLUIDS IN OUR EVERYDAY LIVES TO THE FANTASTICAL PLUMES CAPTURED BY THE HUBBLE TELESCOPE. HOWEVER THE MECHANICS OF TURBULENCE ARE STILL A POORLY UNDERSTOOD UNSOLVED PROBLEM IN CLASSICAL PHYSICS..THE CORE OF MY RESEARCH IS UNDERSTANDING AND QUANTIFYING THE KEYSTONE ROLE THAT FLUID TURBULENCE PLAYS IN VARIOUS ASTROPHYSICAL ENVIRONMENTS SUCH AS THE INTERSTELLAR MEDIUM (ISM). UNDERSTANDING THESE PROCESSES IS VITAL FOR ATTAINING A HOLISTIC DESCRIPTION OF STAR FORMATION AND ULTIMATELY GALAXY EVOLUTION. NOT ONLY ARE SUCH SYSTEMS INCREDIBLY COMPLEX WITH MULTIPLE COMPETING FACTORS CONTRIBUTING TO THE PHENOMENA OBSERVED GIVEN THE UNFATHOMABLE DISTANCES THAT THESE OBJECTS ARE AWAY FROM US TURBULENCE IS ESPECIALLY DIFFICULT TO QUANTIFY OBSERVATIONALLY IN EXTRATERRESTRIAL SYSTEMS. IT IS DIFFICULT IF NOT IMPOSSIBLE TO ATTAIN NEAT ANALYTIC.SOLUTIONS TO SUCH CHAOTIC SYSTEMS INVOLVING MULTIPLE SPATIAL AND TEMPORAL SCALES. HOWEVER FOLLOWING AN INCREDIBLE DECADE IN THE RAPID DEVELOPMENT OF AUTOMATION TOOLS MACHINE LEARNING (ML) TECHNIQUES AND NEURAL NETWORKS (NN) ARE THE STATE OF THE ART WHEN IT COMES TO PATTERN RECOGNITION TECHNOLOGY AND I WILL LEVERAGE THESE ADVANCES TO BRING ABOUT SYMBIOTIC DEVELOPMENT OF BOTH REALMS OF ASTRONOMY AND ML..MY APPROACH IS TO DEVELOP NEW INTERPRETABLE ML TOOLS FOR TURBULENCE THAT SIMULTANEOUSLY ADDRESS CHALLENGES BOTH IN FLUID MECHANICS AND DEEP LEARNING (DL) PRACTICES. A GRAND CHALLENGE IN THE STUDY OF FLUID TURBULENCE IS TO ACHIEVE DETAILED NUMERICAL SIMULATIONS. TO ADDRESS THIS A PORTION OF THIS RESEARCH FOCUSES ON DEVELOPING PHYSICS-INFORMED CONVOLUTIONAL NEURAL NETWORKS FOR SUPER-RESOLUTION ON SIMULATIONS OF TURBULENT RAYLEIGH-BENARD CONVECTION (RBC).