plot of model get clipped

[HaoLi111]

I tried to adjust the boundary, (including letting the plot size to be 'fixed'), but some words are still clipped. Could anyone adjust the plotting function to make the ** default ** boundary larger?

[mnwright]

I haven't looked into the plot function yet since I'm maintainer of the package. It's not easy to customize and definitely needs some refactoring.

Some fine tuning is possible with the arrow.length argument:

library(neuralnet)
nn <- neuralnet(Species ~ ., hidden = c(4, 4, 3), data = iris)

plot(nn, rep = 1) # Clipped
plot(nn, rep = 1, arrow.length = 0.15) # Better

You could also use x.entry and x.out to adjust the positions of the input and output layers. The defaults are:

x.entry <- 0.5 - (arrow.length/2) * length(weights)
x.out <- 0.5 + (arrow.length/2) * length(weights)

where length(weights) is the number of hidden layers plus one.