norm of the discrepancy between the computed approximation UV ∗ and the matrix A being decomposed

[Tagar]

There is an interesting distributed SVD implementation from Criteo :
https://github.com/criteo/Spark-RSVD

Can you please share if you used some particular methods to evaluate norm of the discrepancy between the computed approximation UV ∗ and the matrix A being decomposed
Like in http://tygert.com/spark.pdf

See
criteo/Spark-RSVD#4

cc @tygert

Thank you!!