Multiple-Layer Networks

and

Backpropagation Algorithms

Multiple-Layer Networks
and

Backpropagation Algorithms
Backpropagation is the generalization of the Widrow-Hoff learning rule to
multiple-layer networks and nonlinear differentiable transfer functions.
Input vectors and the corresponding target vectors are used to train a
network until it can approximate a function, associate input vectors with
specific output vectors, or classify input vectors in an appropriate way as
defined by you.

Architecture
This section presents the architecture of the network that is most
commonly used with the backpropagation algorithm
the multilayer feedforward network

Architecture
Neuron Model
An elementary neuron with R inputs is shown below. Each input is
weighted with an appropriate w. The sum of the weighted inputs and the
bias forms the input to the transfer function f. Neurons can use any
differentiable transfer function f to generate their output.

Architecture
Neuron Model
Transfer Functions (Activition Function)
Multilayer networks often use the log-sigmoid transfer function logsig.
The function logsig generates outputs between 0 and 1 as the neuron's
net input goes from negative to positive infinity

Architecture
Neuron Model
Transfer Functions (Activition Function)
Alternatively, multilayer networks can use the tan-sigmoid transfer
function-tansig.
The function logsig generates outputs between -1 and +1 as the neuron's
net input goes from negative to positive infinity

Architecture
Feedforward Network
A single-layer network of S logsig neurons having R inputs is shown
below in full detail on the left and with a layer diagram on the right.

Architecture
Feedforward Network
Feedforward networks often have one or more hidden layers of sigmoid neurons followed
by an output layer of linear neurons.
Multiple layers of neurons with nonlinear transfer functions allow the network to learn
nonlinear and linear relationships between input and output vectors.
The linear output layer lets the network produce values outside the range -1 to +1. On the
other hand, if you want to constrain the outputs of a network (such as between 0 and 1),
then the output layer should use a sigmoid transfer function (such as logsig).

Learning Algorithm:
Backpropagation
The following slides describes teaching process of multi-layer neural network
employing backpropagation algorithm. To illustrate this process the three layer neural
network with two inputs and one output,which is shown in the picture below, is used:

Learning Algorithm:
Backpropagation
Each neuron is composed of two units. First unit adds products of weights coefficients and
input signals. The second unit realise nonlinear function, called neuron transfer (activation)
function. Signal e is adder output signal, and y = f(e) is output signal of nonlinear element.
Signal y is also output signal of neuron.

Learning Algorithm:
Backpropagation
To teach the neural network we need training data set. The training data set consists of
input signals (x1 and x2 ) assigned with corresponding target (desired output) z.
The network training is an iterative process. In each iteration weights coefficients of nodes
are modified using new data from training data set. Modification is calculated using
algorithm described below:
Each teaching step starts with forcing both input signals from training set. After this stage
we can determine output signals values for each neuron in each network layer.

Learning Algorithm:
Backpropagation
Pictures below illustrate how signal is propagating through the network,
Symbols w(xm)n represent weights of connections between network input xm and
neuron n in input layer. Symbols yn represents output signal of neuron n.

Learning Algorithm:
Backpropagation

Learning Algorithm:
Backpropagation

Learning Algorithm:
Backpropagation
Propagation of signals through the hidden layer. Symbols wmn represent weights
of connections between output of neuron m and input of neuron n in the next
layer.

Learning Algorithm:
Backpropagation

Learning Algorithm:
Backpropagation

Learning Algorithm:
Backpropagation
Propagation of signals through the output layer.

Learning Algorithm:
Backpropagation
In the next algorithm step the output signal of the network y is
compared with the desired output value (the target), which is found in
training data set. The difference is called error signal d of output layer
neuron

Learning Algorithm:
Backpropagation
The idea is to propagate error signal d (computed in single teaching step)
back to all neurons, which output signals were input for discussed
neuron.

Learning Algorithm:
Backpropagation
The idea is to propagate error signal d (computed in single teaching step)
back to all neurons, which output signals were input for discussed
neuron.

Learning Algorithm:
Backpropagation
The weights' coefficients wmn used to propagate errors back are equal to
this used during computing output value. Only the direction of data flow
is changed (signals are propagated from output to inputs one after the
other). This technique is used for all network layers. If propagated errors
came from few neurons they are added. The illustration is below:

Learning Algorithm:
Backpropagation
When the error signal for each neuron is computed, the weights
coefficients of each neuron input node may be modified. In formulas
below df(e)/de represents derivative of neuron activation function
(which weights are modified).

Learning Algorithm:
Backpropagation
When the error signal for each neuron is computed, the weights
coefficients of each neuron input node may be modified. In formulas
below df(e)/de represents derivative of neuron activation function
(which weights are modified).

Learning Algorithm:
Backpropagation
When the error signal for each neuron is computed, the weights
coefficients of each neuron input node may be modified. In formulas
below df(e)/de represents derivative of neuron activation function
(which weights are modified).

Inisialisasi jaringan Backpropagation

net= newff(PR,[S1 S2...SN], (TF1 TF2....TFN), BTF,BLF, PF)
Keterangan
Net= jaringan backpro yg terdiri dari n layer
PR= matriks ordo Rx2 berisi nilai min dan max
Si(i=1,2,....n)= jml unit pd layer ke i
TFi(i=1,2,....n)=fungsi aktivasi, default=tansig(sigmoid
bipolar)
BTF= fungsi pelatihan jaringan. Default= traindx
BLF=fungsi perubahan bobot.default= learngdm
PF= fungsi perhitungan error. Default=mse

Latihan 1
Buat inisialisasi backpro yg melatih jaringan yg
terdiri dari 2 masukan, sebuah layer
tersembunyi yg terdiri dari 3 unit dan sebuah
keluaran (2-3-1). Dengan data sbb:
x1

x2

-1

-1

-1

-1

p= [ -1 -1 2 2; 0 5 0 5]
t= [-1 -1 1 1]
net=newff ([-1 2; 0 5], [3 1], {tansig, purelin})

Inisialisasi Bobot
Matlab akan memberikan nilai bobot dan bias
awal dengan bilangan acak kecil.
Bobot dan bias akan berubah setiap kali kita
membentuk jaringan

Contoh 15.2

Lapisan tersembunyi
1

Unit Masukan
x1

x2

bias

z1

-1.3

0.7

0.3

z2

0.5

-0.1

z3

1.3

-0.4

-0.9

z4

-0.1

1.2

0.5

>> net=newff([-1 2;-1 2], [4,3,1])

>> net.IW{1,1}
ans =
-1.4034 1.2308
1.4855 -1.1304
1.8399 -0.3149
-0.0656 1.8655
>> net.b{1}
ans =
2.8863
-1.1109
0.1708
-3.6999

Lapisan
Tersembunyi
2

Lapisan Tersembunyi 1
z1

z2

z3

z4

bias

v1

0.4

0.3

-1

-0.3

0.5

v2

0.6

-0.6

-1.2

-1.3

v3

0.4

-0.3

0.2

0.9

-0.3

>> net.LW{2,1}
ans =
0.9173 -1.4575 0.5611 -0.3383
1.3008 0.9883 0.7296 0.4401
0.4656 1.2975 0.7269 -0.9830
>> net.b{2}
ans =
-1.8425
0
1.8425

Keluaran
y1

Lapisan Tersembunyi 2

v1

v2

v3

bias

0.4

0.9

-0.1

-1

>> net.LW{3,2}
ans =
0.5169 -1.1745 -0.5597
Coba ketikkan
net.IW {2,1}
net.LW{3,1}
net.LW{1,2}
net.IW {2,2}
Apa yang terjadi? Berikan alasan saudara!

 Untuk mengubah nilai bobot dan bias seperti tabel maka

dapat dilakukan sbb:
>> net.IW{1,1}=[-1.3 0.7; 0.5 0; 1.3 -0.4; -0.1 1.2];
>> net.b{1}=[0.3; -0.1; -0.9 ; 0.5];
>> net.LW{2,1}= [0.4 0.3 -1 -0.3; 0.6 0 -0.6 -1.2; 0.4 -0.3 0.2 0.9];
>> net.b{2}=[0.5 ; -1.3 ; -0.3];
>> net.LW{3,2}=[0.4 0.9 -0.1];
>> net.b{3}= [-1];

Simulasi Jaringan
Hitunglah keluaran jaringan contoh 15.2 jika
diberikan masukanx1=0.5 dan x2=1.3.

p=[0.5 ;1.3];
net=newff([-1 2;-1 2], [4,3,1])
net.IW{1,1}=[-1.3 0.7; 0.5 0; 1.3 -0.4; -0.1 1.2];
net.b{1}=[0.3; -0.1; -0.9 ; 0.5];
net.LW{2,1}= [0.4 0.3 -1 -0.3; 0.6 0 -0.6 -1.2; 0.4
-0.3 0.2 0.9];
net.b{2}=[0.5 ; -1.3 ; -0.3];
net.LW{3,2}=[0.4 0.9 -0.1];
net.b{3}= [-1];
y=sim(net,p)

 Jika menginginkan target =1 maka tambahkan

sbb:
t=[1];
[y, Pf,Af,e,perf]=sim (net,p,[],[],t)

Pelatihan backpropagation
Jika diketahui data sbb:
x1

X2

-1

-1

-1

-1

 p=[-1 -1 2 2 ; 0 5 0 5];
t=[-1 -1 1 1];
net=newff (minmax (p), [3,1],
{'tansig','purelin'}, 'traingd');
net.IW{1,1} % melihat bobot awal
net.b{1} % melihat bias awal
net.LW{2,1}% melihat bobot pada lapisan 2
net.b{2} %melihat bobot lapisan 2
[y, Pf,Af,e,perf]=sim (net,p,[],[],t)% simulasi
jaringan
net=train(net,p,t) % melatih jaringan

 Mengetahui bobot setelah dilatih

net.IW{1,1} % melihat bobot awal
net.b{1} % melihat bias awal
net.LW{2,1}% melihat bobot pada lapisan 2
net.b{2} %melihat bobot lapisan 2
[y, Pf,Af,e,perf]=sim (net,p,[],[],t)% simulasi
jaringan

Penambahan parameter pelatihan

p=[-1 -1 2 2 ; 0 5 0 5];
t=[-1 -1 1 1];
net=newff (minmax (p), [3,1], {'tansig','purelin'}, 'traingd');
net=init(net);
net.trainParam.show=100;
net.trainParam.mc=0.5;
net.trainParam.lr=0.1;
net.trainParam.epochs=100;
net.trainParam.goal=0.0001;
net.trainParam.time;
net=train(net,p,t) % melatih jaringan

 Ubahlah nilai tiap paramater, apa kesimpulan

anda?
Ubahlah nilai tiap paramater, apa kesimpulan
anda?

Praktek
Jaringan

12-2-1
12-3-1
12-4-1
12-5-1
12-6-1

Jumlah epoch pd toleransi 10e-5

Fungsi Pelatihan
traingd
traingdm

traingda
traingdx
trainrp
traincgf
traincgp
traincgb

Jumah epoch pd toleransi 10e-5