Gaussian Process Tomography with Deep Neural Networks

Francisco Matos
Max Planck Institute for Plasma Physics
Tokamak Theory

Tomographic reconstruction of the plasma emissivity has several applications in fusion experiments. In the context of plasma diagnostics, Tomography is an inverse problem, usually requiring the usage of algorithms based on regularization to obtain a solution. We propose to solve this problem using a hybrid approach of Gaussian Processes and Deep Neural networks.


Back to Science at Extreme Scales: Where Big Data Meets Large-Scale Computing