FIG. 5.

The performance characteristics of sparse and dense matrices representing the same MSM. (a) The mean number of transitions per state in a transition counts matrix as a function of the number of states in the model. Any pair of states with an observed transition between them has a nonzero entry in the transition counts matrix and consumes memory in both sparse and dense cases. In contrast, a sparse matrix does not require memory for zero elements of the transition counts matrix. (b) The runtime of an eigendecomposition as a function of the number of states in a model. (c) The memory footprint of the transition probability matrix as a function of the number of states in a model.