Figure 4
From: Direct vs 2-stage approaches to structured motif finding
Examples: SISMA_Smile vs Risotto. Running times of Risotto, SISMA_Smile’s stages 1 and 2, and of all SISMA_Smile’s list intersection step (during stage 2). A 0s time for list intersection means that the corresponding step took time smaller than timer resolution. The box index selection order during stage 2 is shown. In example (a) SISMA_Smile outperforms Risotto. Observe that SMILE is called once on the (10,2) pair, so that the time reported for the 6th box is 0. In this example Risotto is 473 times slower than SISMA. In example (b) Risotto outperforms SISMA_Smile because the stage 1 performed by SMILE is slow due to the presence of a box for which a large number of simple motifs is found. In this case SISMA_Smile is 28 times slower than Risotto. In particular, the most time consuming task is the extraction of the (15,4) box (about 91.7% of total execution time), for which 21631 simple motifs are found.