next up previous
Next: 5 Coding Guidelines Up: 4 Experimental Code Previous: 4.2 Hybrid MLNs

4.3 Parallel Alchemy

There are many opportunities to parallelize the algorithms in Alchemy. We have implemented a version of the weight learning algorithms which, in the case of multiple databases, distributes the inference in each database to a different node. This is achieved using the Message Passing Interface (MPI).

Marc Sumner 2008-09-17