Академический Документы
Профессиональный Документы
Культура Документы
neural networks
Abin - Roozgard1
Presentation layout
Introduction
LeNet 5
Comparison
Disadvantage
Application
2
i n t r o d u c t i o n
3
Introduction
4
Drawbacks
5
Behavior of multilayer neural networks
Three-Layer Arbitrary
(Complexity A B
Limited by No. B
A
of Nodes) B A
6
Multi-layer perceptron and image processing
7
Drawbacks of previous neural networks
8
Drawbacks of previous neural networks
9
Drawbacks of previous neural networks
10
Drawbacks of previous neural networks
11
Drawbacks of previous neural networks
12
Drawbacks of previous neural networks
Feature 1
Feature 2
13
Drawbacks of previous neural networks
32 * 32 input image
14
Drawbacks of previous neural networks
A A
15
Improvement
16
Convolutional neural network
(CNN)
17
History
18
About CNN’s
19
About CNN’s
Classification f1
f2
Input
Pre-processing for f1 … fn Classification output
feature extraction
21
CNN’s Topology
Feature maps
C
Feature extraction layer
Convolution layer
22
Feature extraction layer or Convolution layer
23
Feature extraction
-1 0 1
Convolve with Threshold
-1 0 1
-1 0 1
24
Feature extraction
Inputs C S
25
Feature extraction
26
Subsampling layer
27
Subsampling layer
28
Subsampling layer
29
Subsampling layer
30
Subsampling layer
Feature map
31
Up to now …
32
LeNet 5
33
LeNet5
Introduced by LeCun.
raw image of 32 × 32 pixels as input.
34
LeNet5
35
LeNet5
y j (v j ) A tanh( Sv j )
36
LeNet5
+1
+1
84
F1 W1 Y j ( Fi Wij ) 2 , j 0,...,9
F2 W2 i 1
0 Y0
+1
W84
F84
37
LeNet5
38
LeNet5
39
LeNet5
40
C o m p a r i s o n
41
Comparison
42
Comparison
43
Disadvantages
44
Disadvantages
From a memory and capacity standpoint the CNN is
not much bigger than a regular two layer network.
45
Disadvantages
Convolution operation
4 nested loops ( 2 loops on input image & 2 loops on kernel)
46
A p p l i c a t i o n
47
Application
Mike O'Neill
48
Application
49
Application
50
Application
51
Application
52
References
53
Questions
4 8
6
54