280
••••••••••• = booIstrap awregate modal RMS ""'"
o IjddM nocIeS .2 h _ nocIeS , 5 _
nodes
!!
•
2 1 . ' OIliO ! !
1.151 0.6 Q i ~ 1.6 1 0.6 0
Biwa • same day modal
o hlddon nodes ;2 _en nocIeS . 5 _ nodes
III C
2
1 .• 1 0.5° 12 1..6'0.10 ' 2 UIi' 0.11 0
Bumnjucl< • same day model
0 - . nod .. ;2 _
nocIeS ! 5 _on nodes
2: U 1 0.5 0 2: '"
050 ! 2 U 1 0.1 0
Darling· same day model
H. Wilson . F. Recknagel
"'
~ • dioll1bu1lOn 01 RMS errOtS • • ouUl ...
~
0 hldden nodes ,2 _ .... nod8S 5 _
nodes
!:!
b
1
,
2 1.6' 0..6° 12 1.11 0..60 ! 2 '" 000
Blwa • 30 day model
o hldden nod .. ;2 hldden nodes ; 5 _
nodes
d
2: .. I o.l0 j :i l 1.6, "0!2 UIi,"O
Bu' rrinjucl< • 30 day modal
o hIdd .... nodes ,2 hldden nod .. , 5 _
nodes
,
!
2: 1.6, ().f0 i 2 16' O'O ~ 2 151 0.40
DarlIng· 30 day modal
Error at which Training Stopped
Figure 14.7. The effect of training time and the number of hidden layer nodes
on test set RMS error. a Biwa same day model. b Biwa 30 day ahead model. c
Burrinjuck same day model. d Burrinjuck 30 day ahead model. e Darling same
day model. f Darling 30 day ahead model.