Multi: 9: For in A
Multi: 9: For in A
l:
9: For multi layer neural networks shown in ligure below. Perform a one_ step
backpropagation training (forward and backward phases) using the learning raG-is 0.1
assume neurons in the hidden layer and output layer have nonlinear signroid function
( unipola r).
u2
t
I
t
F*fr r.*",1 fo, hl .lJerft L*luy
nd:'= L*jiX; + #i
#l -- Lt's
nelt -- Z.d*l + ()'?i6 I + 1.1
()J -- F tr,*lf ) --
t + AnLll'
Cz E,
,
l+e
@ forrarorJ for D/p tAJw
,{_-_
rterl €w4j Oj
nctt -: {,zl(" o.qffi+ ('o'\ # o.gql) +0,5)ft =l,Lllg
I
o,8l
^l,qlq =
l+e
o*sj =Dsp oj
)
- t,1lq
>t c o.o3l
UJtt *-, \. -
f.
-t {r*l fi c2,o$ t
= *o.j f o*l * a ro 3l -
ll-
tilt'?-
tNli % elL5.*& d.J-* o. o 3 I
ilr
k" h:Jolun-
Ie
wL\- r€,.1+ o.{ 4 ot ooa sqt '*
_r
\i rln v\ L \.1 a-. .-1 ( \ {\'., {>'*\* r j
\ -{r
"}
\ r*') 'r {ta*) s
.--t-\ i'*-' ,^i ,- \,. '. \-(L o* \
\
L ^T
_C
i-
+ icx):
t:
x-\
\- -.te
tJ\
\
/
\ \l rr * Q._,\o 1 )
,,
- (-/' )
Lt -+
qcx) :
) a^(
tJ
eAt [,-.c. lr"n,
x -f' \
L'*e- +F.*)=\ T
u)
Scanned by CamScanner