neural process family (nps, cnps, anps, etc.)

These are a class of models designed to represent distributions over functions, allowing them to be used for tasks like meta-learning and conditional density estimation.