Table 2.

Computational time in seconds

	Algorithm	McCaskill	CONTRAfold
1	γ-centroid with fixed γ	22	47
2	gCentroid-pMCC	36	59
3	Max-pMCC (1000)	178	303
4	Max-pMCC (10000)	1425	2391
5	Max-pMCC (100000)	13910	23291
6	Max-pMCC (1000000)	138987	232397

Total computational time in seconds for predicting secondary structures of all RNA sequences in the S-151Rfam dataset. The 1st row indicates the γ-centroid estimator [10] with a fixed γ parameter (1 for McCaskill model and 2 for CONTRAfold model). The 2nd row indicates the prediction of RNA secondary structure with the γ-centroid estimator and pseudo-expected MCC (Method M2). "Max-pMCC (N)" from the 3rd to 6th rows indicate the estimator of Eq. (23), that is, RNA secondary structure prediction by maximizing pseudo-expected MCC with stochastic sampling (Method M1) where N is the number of samples.