Approximate k-Closest-Pairs in Large High Dimensional Data