none
Does Revo R support Convex curve and non linear constraints? RRS feed

  • Question

  • We are from analytics background and are using SAS for performing non linear optimization.

    We are evaluating if Revo R can be used for optimization problem solving instead of SAS. 

    Our business application supports two types of function convex and S-shaped curves and linear & non-linear constraints. These constraints can be combined with any one type of functional form at a time.

    Example of convex curve –

      Y= c〗^k*pow(a^k,p^k)

    Example of S-shaped curve –

    Y= c〗^k*pow(a^k,p^k )/(b^k+ pow(a^k,p^k )  )

    Example of non-linear constraints –

    Min Bound (50%) < ∑_(k=0)^n▒〖c^k*pow(a^k,p^k)〗 < Max Bound (150%)

    Example of linear constraints –

    Min Bound (50%) < a+b+c < Max Bound (150%)

    At present we are using SAS to solve these business problems. We are looking for SAS replacement software, which can solve similar kind of problems with performance equivalent to SAS. 


    I am looking for following capabilities in Revo R as well –

    1. Can Revo R read data from Hadoop eco system?
    2. Can Revo R be used in distributed computing architecture? E.g. Apache Spark
    3. Please share benchmarking of its performance
    4. How does Revo R perform when number of variables keep on increasing

    Thanks In Advance!

    Friday, November 4, 2016 10:09 AM

All replies

  • There is lots of capability in R for optimization. A high-level summary is available from this R Project Optimization Task View.  Regarding your other questions:

    1. Can Revo R read data from Hadoop eco system? 

    >> Yes. In addition to the facilities in open source R, R Server can stream data from text and XDF files in HDFS directly into our ScaleR algorithms for analysis via Map Reduce or Spark. In our next version, we'll add the ability to read Parquet and Hive data directly into Spark DataFrames for analysis. 

    2. Can Revo R be used in distributed computing architecture? E.g. Apache Spark

    >> Yes, the parallel distributed algorithms in R Server’s ScaleR library run on top of both Map Reduce and Spark. 

    3. Please share benchmarking of its performance

    >> Please contact your Microsoft account team to arrange a discussion on this topic.

    4. How does Revo R perform when number of variables keep on increasing

    >> That is heavily dependent on the data size, algorithm, hardware, etc. Please contact your Microsoft account team to arrange a more detailed discussion on this topic.  

    Tuesday, November 8, 2016 6:16 PM