Deep and Multi-Fidelity Learning with Gaussian Processes

Explore deep and multi-fidelity learning with Gaussian processes in this 25-minute conference talk by Andreas Damianou from Amazon. Gain insights into uncertainty quantification (UQ) and its application in understanding physical systems and decision-making under uncertainty. Learn how Gaussian Process (GP) emulators can replace computationally expensive and complex codes with inexpensive and functionally simple approximations. Discover the theoretical and numerical aspects of GP emulation, with a focus on real-world applications featuring 'large' characteristics, such as complex physical and numerical models or extensive datasets. Delve into specific global challenges, including climate, tsunami, and earthquake problems, to understand how these techniques can be applied to address pressing issues.

Deep and Multi-fidelity learning with Gaussian processes: Andreas Damianou, Amazon