Flexible Aggregate Similarity Search in High-Dimensional Data Sets