by Xiaocheng Tang [https://sourl.cn/TYRu4G]
This code implements distributed L-BFGS using Apache Spark. Compared with the L-BFGS implementation included in Spark, this code also distributes the storage of L-BFGS history and the computations of descent direction, based upon the idea proposed in this 2014 NIPS paper.
The present code trains l2-regularized logistic regression.
How to run the code locally:
sbt/sbt package
# need to update the path to spark-submit
./run-demo-local.sh