We determine the quality of the configurations by the posterior probability of each configuration given the pairwise IBD relationships, or the emission probability of the corresponding HMM state.
The main steps of Metropolis algorithm are (i) to fix the temperature, (ii) to fix the initial spin configuration, (iii) to find the system energy for initial configuration, (iv) to flip arbitrarily one spin from the system, (v) to find the system energy of new configuration, and (vi) to evaluate the transition probability of new configuration.
The main steps of Metropolis algorithm are (i) to fix the temperature, (ii) to fix initial spin configuration, (iii) to find the system energy for initial configuration, (iv) to flip arbitrarily one spin from the system, (v) to find the system energy of new configuration, and (vi) to evaluate the transition probability of new configuration.
The proposed definition is [ 5] (2) where k is a positive constant (which sets the dimension and scale), w is the number of distinct configurations of the system, p i is the probability of such configuration and q ∈ ℛ is the entropic parameter.
To obtain the top- k configurations, a naïve way is to compute the posterior probability of each configuration and then rank them.
In Bayesian statistical inference, the product of the likelihood and the prior probability of the configuration of model parameters is proportional to the desired posterior probability of the configuration.
Thus, it is necessary to sum over all the configurations ξ ¯ for which the channel constraint is satisfied and the j th symbol is in state a and add up the probability of the configuration, as follows T ij a [ n ] = P ξ j = a | { r μ } μ ∈ I ( i ), μ ∈ I ( i ) P r j | ξ j = a · P ξ j = a = ∑ ξ ¯ : ξ j = a P ε i | ξ ¯ ∏ k ∈ I ( i ) ∖ j S ik ξ k [ n − 1 ] (13).
The emission probability is computed as the posterior probability of the configuration given all the pairwise IBD relationships for this position.
The average adaptive value of each strategy (wm, wp and wb) is obtained by computing the sum of the four products (probability of each configuration multiplied by its corresponding fitness, as indicated in Table 1).
