Table 1.

We show the average posterior error $\tilde{P}=\frac{1}{2T}{\sum }_t{\sum }_i|P\left(q_t=i|{\theta }^M,O\right)-P\left(q_t=i|{\theta }^{true},O\right)|$ and total number of mismatches between the two Viterbi paths , generated by models with parameters θ^true and θ^M

Method	w (in σD)	$\tilde{P}$		Likelihood	Time(in sec)	Speed up
	0.25	0.001	3	-5470	74	1.2
	0.50	0.001	3	-5475	61	1.4
	0.75	0.002	6	-5469	35	2.4
Approx	1.00	0.004	22	-5478	21	4.2
	1.25	0.012	81	-5588	13	6.5
	1.50	0.054	410	-6576	8	10.4
	1.75	0.219	2345	-8230	4	20.1
	2.00	0.248	2857	-8492	3	34.1
FBG	...	0.003	12	-5471	87	1.0
True	...	...		-5470		...