Efficient incorporation of model changes into a tree-structured factor graph

Model updates take O(log n) time each

Query marginal distributions in O(log n) time; maintain MAP configurations in O(d log n/d) time, where d is the number of changes in the configuration.

Uses Gaussian process regression to estimate the profile of each gene, while allowing each replicate organism to
shift slightly in time relative to the others, to account for biological variation in development speed.
(:endif:)