Performance benchmarks for varying .

<p>In this test our input data have the dimension and the number of input objects/vectors . <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0074113#pone-0074113-g007" target="_blank">Figure 7(a)</a> shows the performance vs. <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0074113#pone.0074113-Garcia1" target="_blank">[23]</a>. <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0074113#pone-0074113-g007" target="_blank">Figure 7(b)</a> shows the performance vs. <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0074113#pone.0074113-Arefin1" target="_blank">[24]</a>. As increases, the total performance gain asymptotically approaches that of matrix multiplication because for large , this computation dominates.