(⅕) DL ~fails in small data regime where uncertainty estimates are required. Neural Processes (NPs) tackle this by meta-learning a distribution of predictors Interested? Check out our blog yanndubs.github.io/Neural-Process… with @AndrewFoongYK @GordonJo76 *same* NP adapting to diff. GPs: pic.twitter.com/zwshA1Recu