1 paper
Prathamesh Patil, Arpit Jain, Aswanth Krishnan
Performance evaluation in AI systems commonly assumes that random dataset splits produce independent and identically distributed (i.i.d.) subsets. We show that this assumption ofte…