r - Line search fails in training ksvm prob.model -

following invalid probability model large support vector machines using ksvm in r:

i training svm using ksvm kernlab package in r. want use probability model, during sigmoid fitting following error message:

line search fails -1.833726 0.5772808 5.844462e-05 5.839508e-05 -1.795008e-08  -1.794263e-08 -2.096847e-12 

when happens, resulting value of prob.model(m) vector of probabilities, rather expected parameters of sigmoid function fitted on these probabilities. causes error , how can prevent it? searching error message yielded no results.

reproducible example:

load(url('http://roelandvanbeek.nl/files/df.rdata')) ksvm(label~value,df[1:1000],c=10,prob.model=true)->m  prob.model(m) # works should, prints list containing 1 named list  # below, non-working problem, unfortunately takes hour due large # sample size ksvm(label~value,df,c=10,prob.model=true)->m # line search fails   prob.model(m) # vector of values 

looking @ source code, this line throws error.

it's on method .probplatt using newton method optimize function, in case platt's scaling. if check line 3007 though you'll see parameters pertaining method.

one of such parameters minstep minimal numeric step method should keep trying optimize function. see, condition of error in line 3090: if (stepsize < minstep). so, basically, function not converging, when reaching minimum step size.

you can try changing minstep lower values circumvent it. alexandros commented these parameters should in interface.