"multiple CPUs with W-J48 decision trees"

sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
edited June 2019 in Help
Hi...I am not getting good performance with W-J48 in RM 5.3, even with the parallel processing extension installed.  I am running an Intel i5 QC 2.5GHz on a MacBookPro with 8GB memory (and forcing RM to run with 6GB).  I tried even running the algorithm in an Amazon EC2 high performance instance, but it did not take advantage of the additional power.  Is this an issue that was resolved in RM 6?
Tagged:

Answers

  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
    update for anyone who is reading...I moved a sample computation to Amazon EC2 using various large computational instances and have seen good increases in performance (instances are running Windows 2008 Server with RM 5.3 - even tried RM 6.0 Prof. Trial) but they do not take advantage of the computational horsepower of these environments.  Does anyone have experience deploying RM on Amazon EC2 and getting good performance?
  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Hi,

    none of the Weka operators are parallelized, so they only use a single cpu, even if dozens are available. To make use of several cpus you can again install the Parallel extension. That way you can at least parallelize e.g. the X-Validation and calculate several W-J48 instances at the same time.

    Best regards,
    Marius
  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
    Aha.  OK Marius.  Thanks for the help.  Are there any plans to parallelize the Weka extensions in the future?
  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    If the Weka toolkit is parallelized the parallization will probably incorporated into RapidMiner when the Weka extension is updated.

    Best regards,
    Marius
Sign In or Register to comment.