A distributed, easy to use solver library for large scale structured SVMs

Apache Spark

Take advantage of lightning-fast in-memory computation, offered by Apache Spark


Library is based on the recent communication efficient distributed CoCoA framework, internally using the primal-dual BCFW Solver.

SVMstruct compatible

Interface is same as in the widely used SVMstruct

Approximate Inference

Compatible with approximate oracles, and automatic caching for efficiency