A Differentiable Framework for Full and Phaseless Data Inversion Using Neural Implicit Contrast-Source Representation
Researchers have developed a new differentiable framework for data inversion using neural implicit representations. This method parameterizes the contrast source as a continuous neural field, improving accuracy and robustness, especially with noisy measurements. The framework can handle both full and phaseless data inversion and allows for super-resolution inference beyond the training grid. AI
IMPACT Introduces a novel neural network approach for scientific data processing, potentially enhancing accuracy in various computational physics applications.