? stato filtrato un testo allegato il cui set di caratteri non era indicato... Nome: non disponibile URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20131128/8a32f3a1/attachment.pl>
help ANN
2 messages · Giulia Di Lauro, S Ellison
1 day later
I would like to create an artificial neural network with R but I don't know its
parameters jet (number of layers, number of neurons,...).
I downloaded the package ANN and I use the function "ANNGA", but I'm
afraid I haven't really created a neural network. In fact, at the end of the
process I have just this output:
Call:
ANNGA.default(x = input, y = output, design = c(1, 3, 1), population = 100,
mutation = 0.2, crossover = 0.6, maxW = 10, minW = -10, maxGen = 1000,
error = 0.001)
**********************************************************
******************
Mean Squared Error------------------------------> 0.01148523
R2----------------------------------------------> 0.6918387
Number of generation----------------------------> 1001 Weight range at
initialization------------------> [ 10 , -10 ] Weight range resulted from the
optimisation-----> [ 13.58698 , -12.93606 ]
**********************************************************
******************
Well, I would like to know if there is in ANN a function to *create* a neural
network and if not, which package I have to download, Nnet?
What you have done is create an ANN object, print it's summary information and then discard it.
if you want to keep it and use it for something, assign it to a variable; for example:
my.ANN <- ANNGA.default(x = input, y = output, design = c(1, 3, 1), population = 100,
mutation = 0.2, crossover = 0.6, maxW = 10, minW = -10, maxGen = 1000,
error = 0.001)
You can then inspect, print, or predict with the object my.ANN. For example, you can say simply
predict(my.ANN)
to see the output values predicted for the inputs you supplied.
This goes for a lot of R fitting functions; they return objects which are printed and discarded unless you assign them to something.
S Ellison
*******************************************************************
This email and any attachments are confidential. Any use...{{dropped:8}}