Machine learning and nuclear theory (my bias): Why?
- ML tools can help us to speed up the scientific process cycle and hence facilitate discoveries
- Enabling fast emulation for big simulations
- Revealing the information content of measured observables w.r.t. theory
- Identifying crucial experimental data for better constraining theory
- Providing meaningful input to applications and planned measurements
- ML tools can help us to reveal the structure of our models
- Parameter estimation with heterogeneous/multi-scale datasets
- Model reduction
- ML tools can help us to provide predictive capability
- Theoretical results often involve ultraviolet and infrared extrapolations due to Hilbert-space truncations
- Uncertainty quantification essential
- Theoretical models are often applied to entirely new nuclear systems and conditions that are not accessible to experiment