The session is about using GPyTorch, as a modern GPU enabled tool for Gaussian Process based modelling and uncertainty quantification for both small and large datasets.
We aim to showcase GPyTorch as a modern, GPU-enabled tool for Gaussian Process training and uncertainty quantification. The topic is important because many researchers value having accurate error bars on their experiment results. Additionally, it is generally good practice to build AI models that have awareness about what they know and what they don't know. Gaussian Processes offer a built-in, principled mechanism to do exactly this.
Read the full description here on GitHub.
