Table 2.

EM, FBG-sampling, and approximate sampling results for simulated, HBL-2, and Corriel dataset are shown here

Dataset	Method	w	F1-measure	Recall	Precision	Time	Compression	Speed-up	Likelihood
		0.50	0.97 ± 0.04	0.96 ± 0.07	0.98 ± 0.02	27	0.387	2.2
		0.75	0.97 ± 0.04	0.96 ± 0.06	0.98 ± 0.03	16	0.195	3.7
		1.00	0.97 ± 0.05	0.95 ± 0.07	0.98 ± 0.03	10	0.097	5.9
	Approx	1.25	0.96 ± 0.06	0.94 ± 0.09	0.98 ± 0.03	7	0.050	8.8
Simulated		1.50	0.94 ± 0.09	0.92 ± 0.12	0.97 ± 0.07	5	0.031	11.3
		1.75	0.91 ± 0.15	0.89 ± 0.18	0.96 ± 0.12	5	0.023	12.2
		2.00	0.86 ± 0.19	0.85 ± 0.21	0.92 ± 0.19	5	0.018	12.2
	FBG	...	0.97 ± 0.04	0.96 ± 0.05	0.98 ± 0.03	58	...	1.0
	EM prior, ML	...	0.96 ± 0.09	0.97 ± 0.04	0.96 ± 0.11	58	...	...
		1.0	0.85 ± 0.00	0.83 ± 0.00	0.88 ± 0.00	72	0.078	11.3
		2.0	0.87 ± 0.00	0.83 ± 0.00	0.91 ± 0.00	21	0.018	39.3
		3.0	0.89 ± 0.00	0.83 ± 0.00	0.95 ± 0.00	13	0.006	61.8
	Approx	4.0	0.84 ± 0.08	0.77 ± 0.11	0.95 ± 0.01	13	0.003	61.9
		5.0	0.71 ± 0.17	0.60 ± 0.22	0.95 ± 0.01	13	0.002	64.8
		6.0	0.79 ± 0.07	0.69 ± 0.10	0.96 ± 0.01	14	0.002	59.3
		7.0	0.76 ± 0.08	0.64 ± 0.11	0.93 ± 0.01	13	0.001	61.4
HBL-2	FBG	...	0.82 ± 0.00	0.84 ± 0.00	0.80 ± 0.00	810	...	1.0
	EM prior, ML	...	0.65	0.90	0.50	810	...	...	15158
	EM prior, best	...	0.85	0.84	0.86	810	...	...	9616
	EM prior, mean	...	0.76 ± 0.09	0.86 ± 0.03	0.68 ± 0.12	810	...	...	13744
	EM unif, ML	...	0.64	0.90	0.50	810	...	...	15159
	EM unif, best	...	0.86	0.84	0.88	810	...	...	9136
	EM unif, mean	...	0.54 ± 0.24	0.77 ± 0.21	0.46 ± 0.27	810	...	...	13457
GM05296	Approx	2.0	0.96 ± 0.00	1.00 ± 0.00	0.93 ± 0.01	3	0.027	13.0
	FBG	...	0.89 ± 0.01	1.00 ± 0.00	0.81 ± 0.01	40	...	1.0
GM00143	Approx	2.0	0.98 ± 0.00	1.00 ± 0.00	0.96 ± 0.00	3	0.027	13.8
	FBG	...	0.89 ± 0.24	1.00 ± 0.00	0.86 ± 0.26	40	...	1.0

Approximate sampling results are reported for different choices of w. The w value which is closest to the one estimated by the L-method is shown in italic. Width w is reported in σ_D of the corresponding dataset, time is reported in seconds, and compression is defined as $\frac{T^{\prime }}{T}$. For HBL-2, the initial parameter values for EM are sampled from the prior or uniform distributions, and the average (mean), most likely (ML), and best (in terms of F1-measure) performances along with likelihoods are reported.