-Artificial Neural NetworkChapter 2 Basic Model 朝陽科技大學 資訊管理系 李麗華 教授 Introduction to ANN Basic Model 1.
Download ReportTranscript -Artificial Neural NetworkChapter 2 Basic Model 朝陽科技大學 資訊管理系 李麗華 教授 Introduction to ANN Basic Model 1.
Slide 1
-Artificial Neural NetworkChapter 2 Basic Model
朝陽科技大學
資訊管理系
李麗華 教授
Slide 2
Introduction to ANN Basic Model
1. Input layer
2. Hidden layer
3. Output layer
4. Weights
5. Processing Element(PE)
6. Learning
7. Recalling
8. Energy function
朝陽科技大學 李麗華 教授
2
Slide 3
ANN Components (1/4)
1. Input layer:[X1,X2,…..Xn]t , where t means vector transpose.
2. Hidden layer: I j => net j => Y j
3. Output layer:Yj
• Three ways of generating output: normalized,
competitive output, competitive learning
4. Weights :Wij means the connection value between layers
Wij
X1
Y1
X2
‧
‧
‧
‧
‧
‧
‧
‧
‧
Yj
Xn
朝陽科技大學 李麗華 教授
3
Slide 4
ANN Components (2/4)
5. Processing Element(PE)
Ij
(A)Summation Function:
W
Xi
ij
(supervised)
i
or I
j
(X
i
W ij )
(unsupervised)
2
i
(B)Activity Function: net
j
n
I j or net
or net
n
j
I j C net
n
n 1
Ij CIj
n
j
n 1
j
(C)Transfer Function:
1.
2.
3.
Discrete type
Linear type
Non-linear type
朝陽科技大學 李麗華 教授
4
Slide 5
ANN Components (3/4)
6. Learning:
– Based on the ANN model used, learning is to
adjust weights to accommodate a set of training
pattern in the network.
7. Recalling:
– Based on the ANN model used, recalling is to
apply the real data pattern to the trained network
so that the outputs are generated and examined.
朝陽科技大學 李麗華 教授
5
Slide 6
ANN Components (4/4)
8. Energy function:
– Energy function is a verification function which
determines if the network energy has converged to its
minimum. Whenever the energy function approaches
to zero, the network approaches to its optimum
solution.
朝陽科技大學 李麗華 教授
6
Slide 7
Transfer Functions (1/3)
• Discrete type transfer function:
1
Yj=
net j > 0
if
0
net j
1
Ynj=
1
Step function or
perceptron fc.
0
<=0
-1
1
net j > 0
Yn-1j if
0
netj=0
Hopfield-Tank fc.
0
net j<0
-1
1
1
if
Yj =
-1
net j > 0
Signum fc.
net j<=0
朝陽科技大學 李麗華 教授
0
-1
7
Slide 8
Transfer Functions (2/3)
• Discrete type transfer function:
1
Yj =
0
if
netj = 0
-1
Yn-1j
-1
Signum0 fc.
0
net j<0
1
Ynj =
1
net j > 0
-1
1
net j > 0
if
net j = 0
BAM fc.
net j<0
朝陽科技大學 李麗華 教授
0
-1
8
Slide 9
Transfer Functions (3/3)
• Linear type:
Yj = net j
net j
net j > 0
Yj =
if
0
net j <=0
• Nonlinear type transfer function:
1
Yj =
Yj =
1
net
j
net
j
net
Sigmoid function
j
net
j
net
j
Hyperbolic Tangent function
朝陽科技大學 李麗華 教授
9
Slide 10
Energy function
(a) The energy function for supervised network learning:
E=
1
2
T
j
Y
j
2
where E is the energy value
j
E
ΔW= ‧
this is the value for adjusting weight Wij
W ij
(b) The energy function for unsupervised network learning:
E=
ΔW=
1
2
X
i
W ij
2
i
‧
E
W ij
this is the value for adjusting weight Wij
朝陽科技大學 李麗華 教授
10
-Artificial Neural NetworkChapter 2 Basic Model
朝陽科技大學
資訊管理系
李麗華 教授
Slide 2
Introduction to ANN Basic Model
1. Input layer
2. Hidden layer
3. Output layer
4. Weights
5. Processing Element(PE)
6. Learning
7. Recalling
8. Energy function
朝陽科技大學 李麗華 教授
2
Slide 3
ANN Components (1/4)
1. Input layer:[X1,X2,…..Xn]t , where t means vector transpose.
2. Hidden layer: I j => net j => Y j
3. Output layer:Yj
• Three ways of generating output: normalized,
competitive output, competitive learning
4. Weights :Wij means the connection value between layers
Wij
X1
Y1
X2
‧
‧
‧
‧
‧
‧
‧
‧
‧
Yj
Xn
朝陽科技大學 李麗華 教授
3
Slide 4
ANN Components (2/4)
5. Processing Element(PE)
Ij
(A)Summation Function:
W
Xi
ij
(supervised)
i
or I
j
(X
i
W ij )
(unsupervised)
2
i
(B)Activity Function: net
j
n
I j or net
or net
n
j
I j C net
n
n 1
Ij CIj
n
j
n 1
j
(C)Transfer Function:
1.
2.
3.
Discrete type
Linear type
Non-linear type
朝陽科技大學 李麗華 教授
4
Slide 5
ANN Components (3/4)
6. Learning:
– Based on the ANN model used, learning is to
adjust weights to accommodate a set of training
pattern in the network.
7. Recalling:
– Based on the ANN model used, recalling is to
apply the real data pattern to the trained network
so that the outputs are generated and examined.
朝陽科技大學 李麗華 教授
5
Slide 6
ANN Components (4/4)
8. Energy function:
– Energy function is a verification function which
determines if the network energy has converged to its
minimum. Whenever the energy function approaches
to zero, the network approaches to its optimum
solution.
朝陽科技大學 李麗華 教授
6
Slide 7
Transfer Functions (1/3)
• Discrete type transfer function:
1
Yj=
net j > 0
if
0
net j
1
Ynj=
1
Step function or
perceptron fc.
0
<=0
-1
1
net j > 0
Yn-1j if
0
netj=0
Hopfield-Tank fc.
0
net j<0
-1
1
1
if
Yj =
-1
net j > 0
Signum fc.
net j<=0
朝陽科技大學 李麗華 教授
0
-1
7
Slide 8
Transfer Functions (2/3)
• Discrete type transfer function:
1
Yj =
0
if
netj = 0
-1
Yn-1j
-1
Signum0 fc.
0
net j<0
1
Ynj =
1
net j > 0
-1
1
net j > 0
if
net j = 0
BAM fc.
net j<0
朝陽科技大學 李麗華 教授
0
-1
8
Slide 9
Transfer Functions (3/3)
• Linear type:
Yj = net j
net j
net j > 0
Yj =
if
0
net j <=0
• Nonlinear type transfer function:
1
Yj =
Yj =
1
net
j
net
j
net
Sigmoid function
j
net
j
net
j
Hyperbolic Tangent function
朝陽科技大學 李麗華 教授
9
Slide 10
Energy function
(a) The energy function for supervised network learning:
E=
1
2
T
j
Y
j
2
where E is the energy value
j
E
ΔW= ‧
this is the value for adjusting weight Wij
W ij
(b) The energy function for unsupervised network learning:
E=
ΔW=
1
2
X
i
W ij
2
i
‧
E
W ij
this is the value for adjusting weight Wij
朝陽科技大學 李麗華 教授
10