Differentiable imaging connects the physical system with its numerical model, so that the optical hardware and the computational algorithm can be optimized together. Hardware and algorithms can then be co-designed for the real physical system.
Learn more: Differentiable Imaging Project
Physics-aware computing builds physical priors and domain knowledge into reconstruction algorithms, to improve image quality and make the reconstruction more stable. Because they use what we know about the physics, these methods can outperform purely data-driven approaches.
Uncertainty-aware computing develops algorithms that estimate and correct system uncertainties such as misalignment, aberrations and calibration errors. With these methods, imaging still works well when the system is not perfect.