Map-Reduce for Machine Learning on Multicore

This work shows that algorithms that fit the Statistical Query model can be written in a certain "summation form," which allows them to be easily parallelized on multicore computers and shows basically linear speedup with an increasing number of processors.