invalid 'times' argument when running neuralnet

[kencheung2]

i have been running into the following error message with the function neuralnet:

Error in rep(" ", (max(nchar(stepmax), nchar("stepmax")) - nchar(step))) :
invalid 'times' argument

in one of the examples, i set threshold at 0.05 and stepmax at 5e+07. When it achieved the threshold (at steps between 1094000 and 1096000), the error occurred.

the problem seems to occur only when i have more than one layer.

i also had a warning message:

package ‘neuralnet’ was built under R version 3.5.2

wondered if it's related

[mnwright]

Could you please provide a reproducible example?

[kencheung2]

hi thanks for responding. i ran the following code on a data set (10 predictors, n=500 rows)

nNodes = c(10,5) thresh = 0.01 nn = neuralnet(f5, data=DAT, hidden=nNodes, threshold=thresh, stepmax=1e+09)

the code would run fine if i used a single layer (i.e., nNodes has one element). i think the problem may be i set stepmax to be too large. is there a max for stepmax? i recognize you may need to do a real run to find a problem - will try to produce a simulated data set, run it, and submit the code and simulated data set. thanks, ken

…

On Jul 9, 2019, at 6:45 AM, Marvin N. Wright ***@***.***> wrote: Could you please provide a reproducible example? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#33?email_source=notifications&email_token=AHTM23MINOHEM2LIM2EF3PDP6RT37A5CNFSM4H6WWBS2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZP4DTI#issuecomment-509592013>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHTM23IVU7TXXF7SDHVTFLLP6RT37ANCNFSM4H6WWBSQ>.