I don't know, but here is a trick I often use when I'm debugging something in R:

put this line on top of the R script:

Code
debugpath <- paste0("D:/Zorro/Strategy/RWDLexample.rda")
//the *.rda file is a complete memory snapshot. Put it anywhere you like.

Code
neural.train <- function(model, XY) {
  save.image(debugpath)
  #load(debugpath)
  ....


(just add these two lines)
Beware: every neural.train function call will save a complete memory snapshot

Now when I'm debugging, I first clear the R session in RStudio, then source the R file and then execute

load(debugpath)

Then I can check line by line what's happening.