You are on page 1of 12

Neural Networks / /

Neural Networks
: 27 2013, 8,198
Introduction :
..

..


..
! Super-Computer



Model
(
Biological Neural Network ) ..

( ..
ANN) . artificial neural network
ANN *
-
..

Biological Neural Network


-
: -

signal
dendrites -
-
) Analog (
fire
Axon
synapses .



.. ANN
Expert System ,, ANN OO Function



Flow (
ANN )Chart
ANN
..
ANN
) Pattern recognition
( )
( -
ANN Hopfield Nets-
Optimization
traveling salesman problem
!
Classification -
- )
(
.. ) Spam (
ANN
.



Hybrid System Reasoning ..
..
Hello World Artificial neuron
neuron
..

( threshold weight
1.5 ) .
2.5
fire
)( 1 1.5
2.5 ..
Boolean value) .. ( .. fire

-. -
) (
AND .. ) (
:

AND logical operation :

fire 1 1.5
1 2
. ) !( fire
: OR

.. true 1

: XOR

Layer
Hidden Hidden Layer) (
Layer ..

fire 1 1 2
1- 1 0 0.5 ) ..
.. 1 (0 = 1
0 1 1
1 0.5 1 fire
fire . 0.5


) (
Supervised training tranning ..
)
Unsupervised training(
validation ..
..

ANN:



inhibit excite ..
) ( ) (
:


Connectivity of patterns

. ) (
AND Gate




:
set of simple processing unit )1
Patten of Connectivity )2
propagate singal )3
Combining input signal )4
Calculating output signal )5
Learning Rule to Adapt weight )6
: )1

) environment (
.. environment input unit .
pre-processed


Fuzzy gate
.
Pattern of Connectivity : )2
Model of Nets

Model model
layer
..
Backward Model
Model . Model

j . i
Layer

..
. Model

: Feed-Forward

propagate signal : )3
) (
Model
..
Combining input signal : )4

: j

vector notation :

. 0.25
) ( 1 bais
: bais

: 0

. w0 1
Calculating output signal )5
(
Activation ) output unit
0 real 1 Activation Function
1- 1 Discrete) ( 1 1
Sigmoid Function Binary Threshold function Identity function
..
.. Identity
)
(
.. Model
Binary Threshold function
1 0

.

1 0 Sigmoid
:

Learning Rule to Adapt weight )6


Supervised
Unsupervised
Classification Unsupervised


Data Mining

Kohonen NN SOM
Hebb ..
:

Supervised
target output ) (
Delta Rule . Bakpropgation
Delta Rule
Simulating Annealing . Genetic Algorithm
:

Delta Rule
Delta Rule
)
Learning rate(
..
Delta Rule :

target . t o

x beta mu 0.1

.
:
Adaline Rule
Widrow-Hoff Rule
Least Mean Squares (LMS) Rule

Hebb Rule
Hebb
:

learning rate . mu o x
:

:


) ( target output
) (
!!


) (
.. ) (
: Root Mean Square RMS

sum
sum
RMS .
:
ANN

fire ) (
fire .*

..

:
Introduction to Neural Networks for Java, Second Edition
C++ Neural Networks and Fuzzy Logic

You might also like