By how much can closed-loop frameworks accelerate computational materials discovery?
Lance Kavalsky, Vinay I. Hegde, Eric S. Muckley, Matthew S. Johnson, Bryce Meredig, Venkatasubramanian Viswanathan
Abstract
A combination of task automation, calculation runtime improvements, machine learning surrogatization, and sequential learning-guided candidate selection within a closed-loop computational workflow can accelerate materials discovery by up to 20×.
Topics & Concepts
WorkflowComputer scienceTask (project management)AutomationSelection (genetic algorithm)Loop (graph theory)Closed loopArtificial intelligenceMachine learningDistributed computingControl engineeringEngineeringSystems engineeringDatabaseMechanical engineeringMathematicsCombinatoricsMachine Learning in Materials ScienceElectronic and Structural Properties of OxidesX-ray Diffraction in Crystallography