Model-based double clustering for high dimensional data