The LOGO computing paradigm provides a non-map-reduce computing method that allows you to parallelize many traditional machine learning algorithms in a statistically efficient manner.