Research Frontiers in Cycle-Consistent Learning for Molecular Property Transformation
CycleGAN-inspired architectures for learning bidirectional mappings between molecular structures and their computed physicochemical properties without paired training data.