10

Click here to load reader

Exmple neural network

Embed Size (px)

Citation preview

Page 1: Exmple neural network

EXAMPLE NEURAL NETWORK

By.Mat

Page 2: Exmple neural network

Example Neural network

Input

3 2 1 2

3

Output

9 4

1 4

9

X^2

Page 3: Exmple neural network

Example Neural network

1 1.5 2 2.5 3 3.5 4 4.5 51

2

3

4

5

6

7

8

9

Input

Output

Example

Page 4: Exmple neural network

Example Neural network

Input

3 2 1 2

3

Output

9 4

1 4

9

Page 5: Exmple neural network

Example Neural network

Input

3 2 1 2

3

Output

9 4

1 4

9

w1

w2

v2

w1’

w2’

v1’

Z1

Z2

input output

bias bias

v1

Page 6: Exmple neural network

Example Neural network

Input

3 2 1 2

3

Output

9 4

1 4

9

45.154

44.7552

-2.84

0.8996

1.0645

-1.70386’

Z1

Z2

input output

bias bias

-6.194/16

*16

tansig

tansig

logsig

oz

Page 7: Exmple neural network

Example Neural network

a = tansig(n) = (2/(1+exp(-2*n)))-1 logsig(n) = 1 / (1 + exp(-n))

Page 8: Exmple neural network

Example Neural networkInput = 3 ====================

output 9

NormalizationNormalization

3/16= 0.1875 9/16= 0.5625

Bias value is always “one”Z1=(0.1875 *45.154)+(1*-6.194) = 2.2723Z2=(0.1875 *44.7552)+(1*-2.84 = 5.5512

Z1’=tansig(Z1)=0.979Z2’=tansig(Z2)=1.000

Page 9: Exmple neural network

Example Neural networkOz=(0.979*0.8991)+(1*1.0645)+(1*-1.70386)= 0.2412

Output= logsig(oz)=0.5600So.. NN output is 0.5600*16 (denormalization) = 8.96And Real output is 0.5625*16 (denormalization) = 9.00

Error=9.00-8.96 = 0.04;Error limits are determined at the time of training

process, The smaller constraint error on the training process, then NN will produce more accurate output.

but the training process will become more difficult.

Page 10: Exmple neural network

Example Neural network

How to get weight?