Professional Documents
Culture Documents
-Application-
Peter Andras
peter.andras@ncl.ac.uk
www.staff.ncl.ac.uk/peter.andras/lectures
Overview
1. Application principles
2. Problem
3. Neural network solution
Application principles
1
.04
1
.02
Control: 0
.98
2 3 4 5
0
.96
0
.94
0
.92
Problem analysis:
• variables
• modularisation into sub-problems
• objectives
• data collection
Problem
A1 A2 A3 A4
B1 M1 M4 M1a M3
B2 M3 M2 M2 M4
B3 M1a M1 M3 M4
Simple solutions get very complex structure.
Problem
Complex solutions by complex methods:
Variable B
Variable A
Example:
Training set: ~ 75% of the data
Validation set: ~ 10% of the data
Testing set: ~ 5% of the data
Neural network solution
Solution design:
neural network model selection
Example:
|| x − w|| 2
−
f ( x) = e 2a 2
x1
x2
yout
x3 Gaussian neurons
Neural network solution
Example: 4 −
|| x − c k ,1 || 2
4 −
|| x − c k , 2 || 2
= ∑ w ⋅e = ∑ w ⋅e
1 1 2 ( a k ,1 ) 2 2 2 2 ( ak , 2 ) 2
y out k ;y out k
W1, W2 k =1 k =1
|| x − c k , 3 || 2 || x − c k , 4 || 2
4 − 4 −
W3, W4 y 3
= ∑ w ⋅e3 2( ak , 3 ) 2
;y 4
= ∑ w ⋅e 4 2 ( ak , 4 ) 2
out k out k
k =1 k =1
… 4 −
|| x − c k ,19 || 2
4 −
|| x − c k , 20 || 2
= ∑ w ⋅e = ∑ wk20 ⋅ e
19 19 2 ( a k ,19 ) 2 20 2 ( a k , 20 ) 2
y ;y
W19, W20 out
k =1
k out
k =1
Neural network solution
Learning the task from the data:
we apply the learning algorithm to each network from
the solution pool
we use the training data set
Example: x (1) = (0.105 ,0.133 ,0.685 )
y 1out (1) = w11 ⋅ f1 ( x (1)) + w12 ⋅ f 2 ( x (1)) + w31 ⋅ f 3 ( x (1)) + w14 ⋅ f 4 ( x (1))
y 1out (1) = 0.997
E = ( y 1out (1) − y1 (1)) 2 = (0.997 − 0.851 ) 2 = 0.0213
w11, new = w11 − c ⋅ 0.146 ⋅ f1 ( x (1))
...
x (1) = (0.105 ,0.133 ,0.685 )
y 1out (1) = w11 ⋅ f1 ( x (1)) + w12 ⋅ f 2 ( x (1)) + w31 ⋅ f 3 ( x (1)) + w14 ⋅ f 4 ( x (1))
y 1out (1) = 0.847
E = ( y 1out (1) − y1 (1)) 2 = (0.847 − 0.851 ) 2 = 0.000016
w11, new = w11 − c ⋅ 0.004 ⋅ f1 ( x (1))
Neural network solution
Learning the task from the data:
5
2.5 5
0
-2.5 4
1 3
2
3 2
4
1
5
15
5
10 5 5
2.5
5
4 0
0 -2.5
4
1 3 1 3
2 2
3 2 2
3
4 4
1
5 51
7.5
5 5 5
2.5 5 2.5 5 5
2.5
0 0 0
4 4 4
-2.5 -2.5 -2.5
1 3 1 3 1 3
2 2 2
2 3 2 3 2
3
4 4 4
5
1 51 51
Neural network solution