Do you not just want to use linear regression?

Of course it requires a DataFrame-like input but that may be more natural to begin with.

If the data set is small, then putting it on the driver and solving locally with a library is pretty easy.

The Cholesky decomposition above doesn't solve the linear system itself, but helps solve AtAx = Atb, because AtA and Atb are small and so that part can be done locally.

On Thu, Oct 6, 2016 at 6:49 AM Cooper <ahmad.rabani.m@gmail.com> wrote:
I have a system of linear equations in the form of Ax = b to solve in Spark.

A is n by n

b is n by 1

I represent 'A' in the form of IndexedRowMatrix or RowMatrix and 'b' in the
form of DenseMatrix or DenseVector.

How can I solve this system to calculate the 'x' vector?

If the suggested solution isĀ  Cholesky Decomposition
<https://github.com/apache/spark/blob/master/mllib/src/main/scala/org/apache/spark/mllib/linalg/CholeskyDecomposition.scala>
, would you please guide me through doing it as it is not part of the public
API ? For example if the original matrix A is:

1,2,3,4
2,1,5,6
3,5,1,7
4,6,7,1

and b is:

5,6,7,8

What is passed as argument to the "solve" method ?

Any other solution other than inversing 'A' would be very helpful.

--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Solve-system-of-linear-equations-in-Spark-tp27847.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe e-mail: user-unsubscribe@spark.apache.org