Fig. 1From: Using multi-step proposal distribution for improved MCMC convergence in Bayesian network structure learningEdge posterior probabilities calculated from five different MCMC chains. The probabilities are compared to each other when the proposal distribution is the usual one-step version. Numbers in the upper-left and lower-right cornersindicate the amount of overlapping dots, and in the titles, “i vs. j” denotes that edge posterior probabilities calculated from chain i are on the x-axis and those from chain j are on the y-axis. For visualization, a small amount of noise is added to the dots to better separate them from each other. The total number of dots (i.e., possible edges) is N 2=112=121Back to article page