In the world of machine learning, every observation are assumped as Independent and identically distributed (IID), but that is not the case all the time. Thus, the specific evaluation methods should be approached for the dataset of IID violation.
This is the university course, which covered 5 different real dataset such as 1) metal icon prediction in water, 2) bio-signal data (for pain assessment), 3) spatial data (soil water permeability), 4)symmetric pair-input data (protein-protien interaction) and 5) No-signal data.
The main idea behind the course was to evaluate the machine learning model by applying the suitable cross-validation methods as per the nature of the dataset. The cross-validations methods such as, leave-replica-out (data set consiting three replicas), leave-subject-out(for biosignal data), spatial CV( for spatial dataset), modified-version of CV (for symetric input-pair data) and permutation test (for no-sginal data) were applied.