Meta-learning Adaptive Deep Kernel Gaussian Processes for Molecular Property Prediction

Official version Version I consider authoritative

Summary

Meta-learning algorithm using GPs. General in scope but with molecular property prediction as the primary application.

My contribution

I had the idea and wrote a prototype algorithm, but Wenlin did the rest of the work.

Thoughts

(as of 2023-11-24)

I think this is a solid algorithm which could be used for real-life problems, if there is a relevant meta-dataset. I would be hesitant to just take our pre-trained model and apply it on a random new problem.