You are on page 1of 5

Convolutional Neural Network

From Theory to Implementation


Muhammad Rizwan & David V. Anderson

September 30, 2015

0 of 4

Ce
n
t
e
rf
o
rSi
g
n
a
l&I
n
f
o
r
ma
t
i
o
nP

Outline

1 of 4

Motivation

Ce
n
t
e
rf
o
rSi
g
n
a
l&I
n
f
o
r
ma
t
i
o
nP

Tentative

Motivation
Block Diagram
Convolution
Pooling
Fully Connected Layer
Learning Parameters
Variants of CNN
Implementation

2 of 4

Ce
n
t
e
rf
o
rSi
g
n
a
l&I
n
f
o
r
ma
t
i
o
nP

Motivation
Using Deep Neural Nets
Image of Size = 50 50 pixels
Architecture
Input Layer = 50 50 = 2500
Hidden Layer 1 = 1000
Hidden Layer 2 = 100
Output Layer = 10
Number of Parameters to Learn
Total Parameters = 2, 500, 000(min)

3 of 4

Ce
n
t
e
rf
o
rSi
g
n
a
l&I
n
f
o
r
ma
t
i
o
nP

Historical Perspective

Key Dates
First successful Deep Nets.

4 of 4

Ce
n
t
e
rf
o
rSi
g
n
a
l&I
n
f
o
r
ma
t
i
o
nP

You might also like