Asynchronous, parallel pseudo-Gibbs classification of the VF dataset

The cryosectioned visible female dataset is a massive dataset spanning the length of the body in 0.33 mm slices. It is infeasible to compute globally over this dataset. However, even when local computations are considered, the dataset is large enough to merit partitioning. In the case of Gibbs classification, such partitioning is inimical to the goal of Gibbs classification. We discuss a parameterized pseudo-Gibbsian approach to classifying the VF dataset which is stronger than ICM but weaker than a full Gibbs classification. We show how it is implemented in terms of asynchronous MPI.