Compact and efficient linear accelerator phase space modeling via implicit neural representation learning
A compact representation of PSF using sinusoidal implicit representation networks (SIRENs) that enables limitless sampling while preserving dosimetric accuracy is introduced and evaluates.