Structure learning with independent non-identically distributed data