References on data partitioning (cross-validation, train/val/test set construction) when data are non-IID
Posted by Adrian, at stats.stackexchange.com,
Consider a prediction setting in which we are interested in training a regression or classification function $f$ with inputs $X …