Apache Spark MLlib：Java中的OLS回归

Question

I have been looking at the documentation of Apache Spark MLlib, and can't seem to find Linear Regression that uses Ordinary Least Squares (OLS). I only see examples of the numerical based algorithm for Stochastic Gradient Descent (SGD) method of doing Linear Regression. I need the closed form OLS Linear Regression method, not SGD.

I am surprised how hard it is to try to find OLS regression in this, when OLS regression is one of the first most basic of Linear Regression methods.

Answer 1

It is hard to find because Spark doesn't provide an implementation. The rationale is pretty simple. If you have data which requires Spark then the cost of computing an analytical solution will be prohibitive.

If you really need this then RowMatrix provides tallSkinnyQR . It should give you something to work with although it is rather unlikely to be useful in practice.