GLM beta constraints

[greif...@gmail.com]

Do beta constraints work for GLM? I cannot make it work. I am using factor variables and I always get something like:

Got exception 'class java.lang.NullPointerException', with msg 'null' java.lang.NullPointerException
at hex.glm.GLM.objVal(GLM.java:830)
at hex.glm.GLM.init(GLM.java:379)
at hex.glm.GLM$GLMDriver.compute2(GLM.java:772)
at water.H2O$H2OCountedCompleter.compute(H2O.java:1017)
at jsr166y.CountedCompleter.exec(CountedCompleter.java:429)
at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263)
at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:974)
at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1477)
at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104)

[tomas....@gmail.com]

Hi Tomas,

looking at your stack trace, the most likely explanation is your beta constraints had betaGiven and no rho argument. We should really fill in default rho which is all zeros, in the meantime you can either remove betaGiven or add rho with all zeros.

Best,

Tomas

[Tomáš Greif]

Hi Tomas,

yes, you are correct. I got confused thinking that I can use betGiven to set exact coefficients to some predictors. After reading the section for proximal operators it is clear that I had to use min/max instead of beta given (or based on your other response "offset").

Tomas