| Library: | nn |
| See also: | rbftest rbfpredict rbfsave rbftrain rbfinfo rbftrain2 |
| Quantlet: | rbfload | |
| Description: | loads a saved RBF network from a file |
| Usage: | rbfnet = rbfload(name) | |
| Input: | ||
| name | string, name of the file which holds the network | |
| Output: | ||
| rbfnet | composed object (list), RBF network as computed by rbftrain or rbftrain2 | |
library("nn") ; load the library
; training set
randomize(2020)
n = 10
xt = normal(n,2)+#(-1,-1)' | normal(n,2)+#(+1,+1)'
yt =(matrix(n)-1)|matrix(n)
; build the RBF network
clusters = 2
learn = 0.1|0.2|0.1
epochs = 5|5
mMSE = 0.05
activ = 0
rbfnet = rbftrain(xt,yt,clusters,learn,epochs,mMSE,activ)
rbfsave(rbfnet,"rbf_net")
rbfnet2=rbfload("rbf_net")
rbfnet.net.clustersWeights ~ rbfnet2.net.clustersWeights
Contents of t [1,] " An 2 - 2 - 1 RBF-network" [2,] " training epochs: 5 - 5" [3,] " cluster's learning rates: 0.2000 - 0.1000" [4,] " output's learning rate: 0.1000" [5,] " minimum mean squared error: 0.050" [6,] " BINARY sigmoid activation function" [7,] " minimum MSE reached: 0.190974 " Contents of _tmp [1,] 0.78395 0.74043 0.78395 0.74043 [2,] 0.33213 0.36666 0.33213 0.36666