Random Forest Implementation in MapReduce

In this talk, we'll describe some of the technical details of Alpine Forest (a Random Forest implementation). In particular, we'll focus on the challenge of implementing machine learning algorithms like Alpine Forest on a MapReduce framework, particularly with regard to the high-cost of MapReduce iterations and the limitations of memory availability on nodes.

We'll also go over a couple of different approaches of implementing decision tree learners (that form the bases of Alpine Forest) on a distributed system like MapReduce. We'll conclude by going over some details of Spark and how they can help with the next version of Alpine Forest.