+ All Categories
Home > Documents > Artificial Neural Networks - Introduction -

Artificial Neural Networks - Introduction -

Date post: 13-Mar-2016
Category:
Upload: valentina-ovid
View: 61 times
Download: 6 times
Share this document with a friend
Description:
Artificial Neural Networks - Introduction -. Overview. Biological inspiration Artificial neurons and neural networks Why use ANN? ANN Characterization. Biological inspiration. - PowerPoint PPT Presentation
Popular Tags:
28
Artificial Neural Networks - Introduction -
Transcript
Page 1: Artificial Neural Networks -  Introduction -

Artificial Neural Networks- Introduction -

Page 2: Artificial Neural Networks -  Introduction -

Overview

1. Biological inspiration

2. Artificial neurons and neural networks

3. Why use ANN?

4. ANN Characterization

Page 3: Artificial Neural Networks -  Introduction -

Biological inspirationAnimals are able to react adaptively to changes in their external and internal environment, and they use their nervous system to perform these behaviours.

An appropriate model/simulation of the nervous system should be able to produce similar responses and behaviours in artificial systems.

The nervous system is build by relatively simple units, the neurons, so copying their behavior and functionality should be the solution.

Page 4: Artificial Neural Networks -  Introduction -

Action potential

Page 5: Artificial Neural Networks -  Introduction -

Biological inspiration

Page 6: Artificial Neural Networks -  Introduction -

Biological inspiration

synapses

axon dendrites

Page 7: Artificial Neural Networks -  Introduction -

Biological inspiration

Dendrites: Input

Axon: Output

Soma : Cell Body

The information transmission happens at the synapses.

Page 8: Artificial Neural Networks -  Introduction -

Artificial neuronsNeurons work by processing information.

The McCullogh-Pitts model

Inputs

Outputw2

w1

w3

wn

wn-1

. . .

x1

x2

x3

xn-1

xn

y)(;

1

zHyxwzn

iii

Page 9: Artificial Neural Networks -  Introduction -

Artificial neural network (ANN)is a mathematical model or computational model based on biological neural networks Artificial Neural Network consists of neurons arranged in layers Neurons act as parallel processorNeurons are connected with each other vi connection.there are weights associated with connectionsImplementation:

Learningtesting

Page 10: Artificial Neural Networks -  Introduction -

Artificial neural networks

Inputs

Output

An artificial neural network is composed of many artificial neurons that are linked together according to a specific network architecture. The objective of the neural network is to transform the inputs into meaningful outputs.

Page 11: Artificial Neural Networks -  Introduction -

Artificial neural networksDendrites: Input LayerAxon : Output LayerSoma: Net( weighted sum of input y_in) and activation functionSynapse: Weights

Page 12: Artificial Neural Networks -  Introduction -

Why use ANN?-Adaptive learning: An ability to learn how to do tasks based on

the data given for training or initial experience. -Self-Organization: An ANN can create its own organization or

representation of the information it receives during learning time.

-Real Time Operation: ANN computations may be carried out in parallel, and special hardware devices are being designed and

manufactured which take advantage of this capability. -Fault Tolerance via Redundant Information Coding: Partial

destruction of a network leads to the corresponding degradation of performance. However, some network capabilities may be

retained even with major network damage.

Page 13: Artificial Neural Networks -  Introduction -

ANN CharacterizationANN can be characterized by:

Activation functionWeights Adjustment (learning algorithm)Architecture

Page 14: Artificial Neural Networks -  Introduction -

Activation functionFunction to map weighted sum of input into outputDetermine whether neuron fires or notLinear function (Identity)

F(y_in)=y y_in (weighted sum of input) y (output)

Step functiony=f(y_in)=

00_10_

yinyyiny

Page 15: Artificial Neural Networks -  Introduction -

Activation functionLogistic or sigmoid function

Binary sigmoid• F(y_in)=

Bipolar Sigmoid• F(y_in)=

)_exp(11

inay

)_exp(1)_exp(1

inayinay

Page 16: Artificial Neural Networks -  Introduction -

Learning AlgorithmLearning in ANN is Weights adjustment to

get desired outputTo minimize the errorTo gain more experience

LearningSupervisedunsupervised

Page 17: Artificial Neural Networks -  Introduction -

Supervised Learning There is supervisor during learning processInput and output are knownThe job of ANN is to classify any new input according to known classesExample : teaching baby the difference pens and other thingsLVQ (learning vector quantization)

Page 18: Artificial Neural Networks -  Introduction -

Unsupervised learningInput known but output unknownThe classes are unknown to ANNJob of ANN is to find similarities between input and divide them into categories (cluster)SOM (Self organizing map)

Page 19: Artificial Neural Networks -  Introduction -

ArchitectureShow the number of layer in Neural NetworkShow the number of neurons in each layerShow how neurons connect to each other

Page 20: Artificial Neural Networks -  Introduction -

ArchitectureFeed forward

allow signals to travel one way only; from input to output. There is no feedback (loops)

Multi layer

Page 21: Artificial Neural Networks -  Introduction -

ArchitectureFeedback networks

signals travelling in both directions by introducing loops in the network

Page 22: Artificial Neural Networks -  Introduction -

linearly separable problemA linearly separable problem is one in which the classes can be separated by a single hyperplane

It is often the case that a problem is not linearly separable. To solve these we use a Multi-Layer Perceptron (MLP) where one layer feeds into the next.

Page 23: Artificial Neural Networks -  Introduction -
Page 24: Artificial Neural Networks -  Introduction -

XOR problem

Page 25: Artificial Neural Networks -  Introduction -

XOR problem

Page 26: Artificial Neural Networks -  Introduction -

The shape of regions in pattern space that can be separated by a Multi-Layer Perceptron

Page 27: Artificial Neural Networks -  Introduction -

back-propagationLast time we saw that the delta rule can be used to train a perceptron. When training the MLP, the error (delta) must be propagated back through the layers. This is called error back-propagation. Or just backpropagation.

Page 28: Artificial Neural Networks -  Introduction -

Recommended