In this recipe, we will demonstrate the use of Spark 2.0's LinearRegression() API to showcase a unified/parameterized API to tackle the linear regression in a comprehensive way capable of extension without backward-compatibility issues of an RDD-based named API. We show how to use the setSolver() to set the optimization method to first-order memory-efficient L-BFGS, which can deal with numerous amount of parameters (that is, especially in sparse configuration) with ease.
In this recipe, the .setSolver() is set to lbgfs, which makes the L-BFGS (see RDD-based regression for more detail) the selected optimization method. The .setElasticNetParam() is not set, so the default of 0 remains ...
