I'm having some trouble with the nnet. Everytime I run it, it comes up with slightly different answers. I assume this is due to using a some sort of random seed to calculate the weights. Doesn't anyone have any examples of how to set these weights so that you can get consistent results? Thanks, Tom Paska
If "nnet" uses standard R random number generators, then calling "set.seed" immediately before "nnet" should produce consistent results. hope this helps. spencer graves Tom Paska wrote:> I'm having some trouble with the nnet. Everytime I > run it, it comes up with slightly different answers. > I assume this is due to using a some sort of random > seed to calculate the weights. Doesn't anyone have > any examples of how to set these weights so that you > can get consistent results? > > Thanks, > Tom Paska > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://www.stat.math.ethz.ch/mailman/listinfo/r-help
Tom Paska wrote:> > I'm having some trouble with the nnet. Everytime I > run it, it comes up with slightly different answers. > I assume this is due to using a some sort of random > seed to calculate the weights. Doesn't anyone have > any examples of how to set these weights so that you > can get consistent results?I guess you have already read the references given in ?nnet: Ripley, B. D. (1996) Pattern Recognition and Neural Networks. Cambridge. Venables, W. N. and Ripley, B. D. (2002) Modern Applied Statistics with S. Fourth edition. Springer. If so, what about specifying some more details so that people can help? It's also a good idea to ask the package authors / maintainers for help, since he / she regularly knows the answer. Uwe Ligges