What is AM-SciML and Scientific Machine Learning?

Machine Learning systems are computational models that learn from data to accurately represent complex real-world systems with unmatched capabilities of analysis, prediction, generation, and control. Think large language models, image generation, and self-driving cars.

Scientific Computing develops mathematical and computational techniques to model challenging problems from Science and Engineering. Think about large-scale computations to model the aerodynamics of airplanes or the flow of blood in the human body; or fast parallel training algorithms for neural networks on GPUs.

Scientific Machine Learning develops Machine Learning models for problems in Science, Medicine, and Technology. It also encompasses the use of Scientific Computing techniques to develop more performant Machine Learning models. Scientific Machine Learning models learn from data while incorporating the rules of physics, biology, and engineering. This makes them more accurate, interpretable, and efficient for solving real-world problems. For example, deep neural network models now often challenge the best known classical mathematical/computational models for describing physical phenomena in the real world.