Sampe code fails in setHyperPars2

[annette987]

I have been trying to run the code you provided in file fouodo-koenig-weihs-ziegler-wright.R
As soon as it gets to the first benchmarking step:

tunwrp.gm.reg <- tuneWrapperGammaMu(type = "regression",
                                    kernel = "lin_kernel", 
                                    opt.meth = "quadprog", method = "CV",
                                    iters.rep = 5)
bench.vet.ssvm.reg <- benchmark(learners = tunwrp.gm.reg, tasks = veteran.task,
                                resamplings = outer, measures = list(c.i, lgrk, hr))

it fails as follows:

[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  0.0009766 is not feasible for parameter 'gamma.mu'!

The full traceback is:

Task: veteran.adj, Learner: surv.survivalsvm.preproc.tuned
Resampling: cross-validation
Measures:             ci        logrank   hr        
[Tune] Started tuning learner surv.survivalsvm.preproc for parameter set:
             Type len Def                                   Constr Req Tunable Trafo
gamma.mu discrete   -   - 0.0009765625,0.001953125,0.00390625,0...   -    TRUE     -
center    logical   -   -                                        -   -    TRUE     -
scale     logical   -   -                                        -   -    TRUE     -
With control class: TuneControlGrid
Imputation value: -0
[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  0.0009766 is not feasible for parameter 'gamma.mu'!

[Tune-x] 1: gamma.mu=0.0009765625; center=TRUE; scale=TRUE
[Tune-y] 1: ci.test.mean=      NA; time: 0.0 min
[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  0.001953 is not feasible for parameter 'gamma.mu'!

[Tune-x] 2: gamma.mu=0.001953125; center=TRUE; scale=TRUE
[Tune-y] 2: ci.test.mean=      NA; time: 0.0 min
[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  0.003906 is not feasible for parameter 'gamma.mu'!

[Tune-x] 3: gamma.mu=0.00390625; center=TRUE; scale=TRUE
[Tune-y] 3: ci.test.mean=      NA; time: 0.0 min
[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  0.007812 is not feasible for parameter 'gamma.mu'!

[Tune-x] 4: gamma.mu=0.0078125; center=TRUE; scale=TRUE
[Tune-y] 4: ci.test.mean=      NA; time: 0.0 min
[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  0.01562 is not feasible for parameter 'gamma.mu'!

[Tune-x] 5: gamma.mu=0.015625; center=TRUE; scale=TRUE
[Tune-y] 5: ci.test.mean=      NA; time: 0.0 min
[Tune-x] 6: gamma.mu=0.03125; center=TRUE; scale=TRUE
Warning in train(.learner$next.learner, .task) :
  Could not train learner surv.survivalsvm: Error in quadprog::solve.QP(C, -d, t(H), f, meq = meq) : 
  constraints are inconsistent, no solution!

Warning in train(.learner$next.learner, .task) :
  Could not train learner surv.survivalsvm: Error in quadprog::solve.QP(C, -d, t(H), f, meq = meq) : 
  constraints are inconsistent, no solution!

[Tune-y] 6: ci.test.mean=     NaN; time: 0.0 min
[Tune-x] 7: gamma.mu=0.0625; center=TRUE; scale=TRUE
Warning in train(.learner$next.learner, .task) :
  Could not train learner surv.survivalsvm: Error in quadprog::solve.QP(C, -d, t(H), f, meq = meq) : 
  constraints are inconsistent, no solution!

Warning in train(.learner$next.learner, .task) :
  Could not train learner surv.survivalsvm: Error in quadprog::solve.QP(C, -d, t(H), f, meq = meq) : 
  constraints are inconsistent, no solution!

[Tune-y] 7: ci.test.mean=     NaN; time: 0.0 min
[Tune-x] 8: gamma.mu=0.125; center=TRUE; scale=TRUE
[Tune-y] 8: ci.test.mean=0.7070195; time: 0.0 min
[Tune-x] 9: gamma.mu=0.25; center=TRUE; scale=TRUE
[Tune-y] 9: ci.test.mean=0.7143521; time: 0.0 min
[Tune-x] 10: gamma.mu=0.5; center=TRUE; scale=TRUE
[Tune-y] 10: ci.test.mean=0.7081605; time: 0.0 min
[Tune-x] 11: gamma.mu=1; center=TRUE; scale=TRUE
[Tune-y] 11: ci.test.mean=0.7264289; time: 0.0 min
[Tune-x] 12: gamma.mu=2; center=TRUE; scale=TRUE
[Tune-y] 12: ci.test.mean=0.7191989; time: 0.0 min
[Tune-x] 13: gamma.mu=4; center=TRUE; scale=TRUE
[Tune-y] 13: ci.test.mean=0.7137441; time: 0.0 min
[Tune-x] 14: gamma.mu=8; center=TRUE; scale=TRUE
[Tune-y] 14: ci.test.mean=0.7136500; time: 0.0 min
[Tune-x] 15: gamma.mu=16; center=TRUE; scale=TRUE
[Tune-y] 15: ci.test.mean=0.6847218; time: 0.0 min
[Tune-x] 16: gamma.mu=32; center=TRUE; scale=TRUE
[Tune-y] 16: ci.test.mean=0.6745995; time: 0.0 min
[Tune-x] Setting hyperpars failed: Error in setHyperPars2.Learner(learner$next.learner, par.vals = par.vals[i]) : 
  64 is not feasible for parameter 'gamma.mu'! 

and so on. Each step fails in turn. Am I doing something wrong?
Thanks.
`

[fouodo]

Hi,

Thanks for this issue. What going wrong is the nonfeasible value of "gamma.mu". Regarding to the "makeRLearner.surv.survivalsvm" from the file "fouodo-koenig-weihs-ziegler-wright.R" parameter "gamma.mu" should take its values in 2^-5 - 2^5. So, just set new values to fix the error.

Regards!