Using Prior-Guided Deep Gaussian Processes

Starting at 9:30 IST

Talk: Accurate and Uncertainty-Aware Multi-Task Prediction of HEA Properties Using Prior-Guided Deep Gaussian Processes by Ahnaf Alvi

Description: This talk presents a novel way to combine parametric and non-parametric models through prior injection. In this talk, a hybrid Encoder-Decoder and Deep Gaussian Process based Multi-Property model is presented. The proposed model can handle heterotopic data, has uncertainty quantification and outperforms other models. The model can be useful as surrogates in discovery campaigns of Self driving labs.

Speaker Bio: The presenter is a third year PhD student in Texas A&M university (TAMU). His doctoral research is funded by a joint TAMU-Los Alamos National Laboratory initiative. Previously he completed his undergraduate degree in mechanical engineering from Bangladesh University of Engineering and Technology.

Join live here: meet.google.com/kuw-jmcf-msg