SlideShare a Scribd company logo
1 of 18
Download to read offline
5/7/2012




  WHAT ARE NEURAL NETWORKS?
 Neural networks are parallel information processing
 systems with their architecture inspired by the
 structure and functioning of the brain

 A neural network have the ability to learn and
 generalize.

 Neural     networks can be trained to make
 classifications and predictions based on historical data

 Neural nets are included in many data mining
 products



                                                                  1
5/7/2012




 Very popular and effective techniques

 Artificial Neural Network goes by many names, such
 as connectionism, parallel distributed processing,
 neuro-computing and natural intelligent systems.

 It is abbreviated by ANN

 ANN has a strong similarity to the biological brain.
 It is composed of interconnected elements called
 neurons.
                   3




The Biological Neuron
  Biological neuron receives inputs from other sources,
   combines them in some way, performs a generally
   nonlinear operation on the result, and then output the
   final result.
                         send Inputs                                    Process the inputs
                         (eyes, ears)




                                        Turn the processed
                                        inputs into outputs

                                                        The electrochemical contact between
                                                        neurons
                                                                                             4




                                                                                                       2
5/7/2012




The Artificial Neuron




             5




Network Layers
 The neurons are
 grouped into layers.

 The input layer.


 The output layer.


 Hidden layers between
 these two layers.

                          6




                                    3
5/7/2012




Neural Network Architecture
 In a Neural Network, neurons are grouped into
  layers.
 The neurons in each layer are the same type.
 There are different types of Layers.
 The Input layer, consists of neurons that receive
   input from the external environment.
 The Output layer, consists of neurons that
   communicate to the user or external environment.
 The Hidden layer, consists of neurons that ONLY
   communicate with other layers of the network.




    Now that we have a model for an artificial
     neuron, we can imagine connecting many of
     then together to form an Artificial Neural
     Network:


    Output layer


    Hidden layer


    Input layer




                                                            4
5/7/2012




  Based on the following assumptions:
1. Information   processing occurs at many simple
 processing elements called neurons.
2. Signals are passed between neurons over
 interconnection links.
3. Each interconnection link has an associated
 weight.
4. Each neuron applies an activation function to
 determine its output signal.




             A Neuron
    x0
                                   - mk
                w0

     x1         w1
                                         f       output y
     xn         wn


   Input      weight    weighted     Activation
  vector x   vector w     sum         function

    The n-dimensional input vector x is mapped into
     variable y by means of the scalar product and a
     nonlinear function mapping




                                                                   5
5/7/2012




What is a neuron?
  a (biological) neuron is a
   node that has many inputs
   and one output                             f
  inputs come from other
   neurons
  the inputs are weighted
  weights can be both positive
   and negative
  inputs are summed at the
   node to produce an
   activation value




The Artificial Neuron Model
 In order to simulate neurons on a computer, we
 need a mathematical model of this node

 Node j ( a neuron) has n inputs xi , i = 1 to n

 Each input (connection) is associated with a weight
 wij

 The neuron includes a bias denoted by bj

 The bias has the effect of increasing or deceasing
 the net input
                 12




                                                              6
5/7/2012




 The net input to node j (net result) ,uj , is the sum of
  the products of the connection inputs and their
  weights:             n
                   u j   w ijxi  b j
                         i 1
Where uj is called the “activation of the neuron”.

 The output of node j is determined by applying a non-
  linear transfer function g to the net input:
                        x j  g(u j )

 called the “transfer function”.




 A common choice for the transfer function is the
  sigmoid:
                                        1
                    g(u j )                u j
                                 1 e

 The sigmoid has similar non-linear properties to the
  transfer function of real neurons:
   It accepts inputs varying from – ∞ to ∞
   bounded below by 0
   bounded above by 1




                                                                   7
5/7/2012




An Example of multilayer feed
forward neural network




                                      8
5/7/2012




Architecture of ANN
 Feed-Forward Networks
  The signals travel one way from input to output


 Feed-Back Networks
  The signals travel as loops in the network, the
   output of the network is connected to the
   input.



            17




Learning ANN
 The purpose of the learning function is to
 modify the variable connection weights
 causes the network to learn the solution to
 a problem according to some neural based
 algorithm.

 There are tow types of learning
    Supervised (Reinforcement) learning.
    Unsupervised learning.

            18




                                                          9
5/7/2012




Supervised learning
 Means their exist an external
 help or a teacher.

 The teacher may be a training
 set of data.

 The target is to minimize the
 error between the desired and
 the actual output.

 The process take place as
 follows:


                     19




   Presenting input and output data to the network,
   this data is often referred to as the “training set”.

   The network processes the inputs and compares
   its resulting outputs against the desired outputs.

   Errors are then propagated back through the
   system, causing the system to adjust the weights,
   which are usually randomly set to begin with.

   This process occurs over and over as a closer
   match between the desired and the predicted
   output.

                20




                                                                10
5/7/2012




 When no further learning is necessary, the
  weights are typically frozen for the
  application.


 There are many algorithms used to implement
  the adaptive feedback required to adjust the
  weights during training. The most common
  technique is called “Back-Propagation”.




Let us suppose that a sufficiently large set of
examples (training set) is available.
Supervised learning:
– The network answer to each input pattern is
directly compared with the desired answer and a
feedback is given to the network to correct possible
errors




                                                            11
5/7/2012




Back Propagation (BP)
   It is an improving performance method in training of
    multilayered feed forward neural networks.

 Are used to adjust the weights and biases of
  networks to minimize the sum squared error of
  the network which is given by
                        1
                SSE 
                        2
                           (xt  xt )2
                                  ˆ

 where xt and xt^ are the desired and predicted
  output of the tth output node.


                                                           23




BP Network – Supervised Training
 Desired output of the training examples

 Error = difference between actual & desired output

 Change weight relative to error size

 Calculate output layer error , then propagate back to
  previous layer

 Hidden weights updated

 Improved performance
                                                           24




                                                                     12
5/7/2012




 Probably the most common type of ANN used
 today is a multilayer feed forward network
 trained using back-propagation (BP)

 Often called a Multilayer Perceptron (MLP)




  Unsupervised learning
 The training set consists of
 input training patterns but not
 with desired outputs.


 This method often referred to
 as self-organization.

 Therefore, the network is
 trained without benefit of any
 teacher.


                   26




                                                    13
5/7/2012




  Applications in Clustering and reducing dimensionality

  Learning may be very slow

  No help from the outside

  No training data, no information available on the
  desired output

  Learning by doing

  Used to pick out structure in the input:
      Clustering
      Compression
                                                          27




Choosing the network size
 It seems better to start with a small number of
  neurons, because:
        learning is faster.
        it is often enough.
        it avoids over-fitting problems.

 If the number of neurons are too much we will get
 an over fit.

 In principle, one hidden layer is sufficient to solve
 any problem. In practice, it may happen that two
 hidden layers with a small number of neurons may
 work better (and/or learn faster) than a network
 with a single layer.
                  28




                                                                    14
5/7/2012




• Too few nodes: Don’t fit the curve very well


• Too many nodes: Over parameterization
  • May fit noise
  • makes network more difficult to train




The Learning Rate
 The learning rate c, which determines by how
  much we change the weights w at each step.

 If c is too small, the algorithm will take a long
  time to converge.
                         Sum-Square Error




                                            Error Surface




                                                            Epoch

                    30




                                                                         15
5/7/2012




 If c is too large, the network may not be
 able to make the fine discriminations
 possible with a system that learns more
 slowly. The algorithm diverges.


                   Sum-Square Error




                                           Error Surface


                                                                      Epoch




                                      31




       Multi-Layer Perceptron

Output vector

                                                           Errj  O j (1  O j ) Errk w jk
Output nodes                                                                       k

                                                                     j   j  (l) Errj
                                                               wij  wij  (l ) Errj Oi
Hidden nodes
                                                            Errj  O j (1  O j )(T j  O j )
                                                              wij                 1
                                                                        Oj           I
                                                                              1 e j
Input nodes
                                                                      I j   wij Oi   j
                                                                              i
Input vector: xi




                                                                                                     16
5/7/2012




Applications of ANNs
 Prediction – weather, stocks, disease, Predicting
 financial time series


 Classification – financial risk assessment, image
 processing

 Data Association – Text Recognition

 Data Conceptualization – Customer purchasing
 habits

 Filtering – Normalizing telephone signals
 Optimization

                                                      33




 Diagnosing medical conditions


 Identifying clusters in customer databases


 Identifying fraudulent credit card transactions


 Hand-written character recognition


 and many more….




                                                                17
5/7/2012




  Advantages
    Adapt to unknown situations
    Autonomous learning and generalization
    Their most important advantage is in solving
     problems that are too complex for conventional
     technologies, problems that do not have an
     algorithmic solution or for which an algorithmic
     solution is too complex to be found.



  Disadvantages
    Not exact
    Large complexity of the network structure


                                                           35




Using a neural network for prediction
  Identify input and outputs
  Preprocess inputs - often scale to the range [0,1]
  Choose an ANN architecture
  Train the ANN with a representative set of training
   examples (usually using BP)
  Test the ANN with another set of known examples
     often the known data set is divided in to training
      and test sets. Cross-validation is a more rigorous
      validation procedure.
  Apply the model to unknown input data




                                                                     18

More Related Content

What's hot

(Artificial) Neural Network
(Artificial) Neural Network(Artificial) Neural Network
(Artificial) Neural Network
Putri Wikie
 
Neural networks
Neural networksNeural networks
Neural networks
Slideshare
 
Neural networks
Neural networksNeural networks
Neural networks
Slideshare
 
Ppt on artifishail intelligence
Ppt on artifishail intelligencePpt on artifishail intelligence
Ppt on artifishail intelligence
snehal_gongle
 

What's hot (20)

(Artificial) Neural Network
(Artificial) Neural Network(Artificial) Neural Network
(Artificial) Neural Network
 
Artificial Neuron network
Artificial Neuron network Artificial Neuron network
Artificial Neuron network
 
Neural networks
Neural networksNeural networks
Neural networks
 
Neural network and mlp
Neural network and mlpNeural network and mlp
Neural network and mlp
 
Neural networks
Neural networksNeural networks
Neural networks
 
Neural network 20161210_jintaekseo
Neural network 20161210_jintaekseoNeural network 20161210_jintaekseo
Neural network 20161210_jintaekseo
 
Artificial neural networks
Artificial neural networks Artificial neural networks
Artificial neural networks
 
Artificial neural network
Artificial neural networkArtificial neural network
Artificial neural network
 
Artificial Neural Network
Artificial Neural NetworkArtificial Neural Network
Artificial Neural Network
 
hopfield neural network
hopfield neural networkhopfield neural network
hopfield neural network
 
Neural network
Neural networkNeural network
Neural network
 
Dr. kiani artificial neural network lecture 1
Dr. kiani artificial neural network lecture 1Dr. kiani artificial neural network lecture 1
Dr. kiani artificial neural network lecture 1
 
Artificial neural network paper
Artificial neural network paperArtificial neural network paper
Artificial neural network paper
 
Neural network
Neural networkNeural network
Neural network
 
Artificial nueral network slideshare
Artificial nueral network slideshareArtificial nueral network slideshare
Artificial nueral network slideshare
 
Artificial Neural Network Paper Presentation
Artificial Neural Network Paper PresentationArtificial Neural Network Paper Presentation
Artificial Neural Network Paper Presentation
 
Artificial Neural Network report
Artificial Neural Network reportArtificial Neural Network report
Artificial Neural Network report
 
Project presentation
Project presentationProject presentation
Project presentation
 
Question bank soft computing
Question bank   soft computingQuestion bank   soft computing
Question bank soft computing
 
Ppt on artifishail intelligence
Ppt on artifishail intelligencePpt on artifishail intelligence
Ppt on artifishail intelligence
 

Viewers also liked (6)

Decision tree for Predictive Modeling
Decision tree for Predictive ModelingDecision tree for Predictive Modeling
Decision tree for Predictive Modeling
 
Data mining techniques and dss
Data mining techniques and dssData mining techniques and dss
Data mining techniques and dss
 
Predictive Modelling
Predictive ModellingPredictive Modelling
Predictive Modelling
 
Discriminant Analysis-lecture 8
Discriminant Analysis-lecture 8Discriminant Analysis-lecture 8
Discriminant Analysis-lecture 8
 
Data mining technique (decision tree)
Data mining technique (decision tree)Data mining technique (decision tree)
Data mining technique (decision tree)
 
Decision tree and random forest
Decision tree and random forestDecision tree and random forest
Decision tree and random forest
 

Similar to ANN-lecture9

Soft Computing-173101
Soft Computing-173101Soft Computing-173101
Soft Computing-173101
AMIT KUMAR
 
Artificial neural network
Artificial neural networkArtificial neural network
Artificial neural network
DEEPASHRI HK
 

Similar to ANN-lecture9 (20)

Neural Networks.pptx
Neural Networks.pptxNeural Networks.pptx
Neural Networks.pptx
 
Soft Computing-173101
Soft Computing-173101Soft Computing-173101
Soft Computing-173101
 
Artificial Neural Networks ppt.pptx for final sem cse
Artificial Neural Networks  ppt.pptx for final sem cseArtificial Neural Networks  ppt.pptx for final sem cse
Artificial Neural Networks ppt.pptx for final sem cse
 
Neural networks of artificial intelligence
Neural networks of artificial  intelligenceNeural networks of artificial  intelligence
Neural networks of artificial intelligence
 
Dr. Syed Muhammad Ali Tirmizi - Special topics in finance lec 13
Dr. Syed Muhammad Ali Tirmizi - Special topics in finance   lec 13Dr. Syed Muhammad Ali Tirmizi - Special topics in finance   lec 13
Dr. Syed Muhammad Ali Tirmizi - Special topics in finance lec 13
 
Acem neuralnetworks
Acem neuralnetworksAcem neuralnetworks
Acem neuralnetworks
 
Artificial neural networks
Artificial neural networks Artificial neural networks
Artificial neural networks
 
Artificial Neural Network
Artificial Neural NetworkArtificial Neural Network
Artificial Neural Network
 
Artificial neural network
Artificial neural networkArtificial neural network
Artificial neural network
 
Basics of Artificial Neural Network
Basics of Artificial Neural Network Basics of Artificial Neural Network
Basics of Artificial Neural Network
 
Artificial Neural Networks for NIU session 2016 17
Artificial Neural Networks for NIU session 2016 17 Artificial Neural Networks for NIU session 2016 17
Artificial Neural Networks for NIU session 2016 17
 
NEURALNETWORKS_DM_SOWMYAJYOTHI.pdf
NEURALNETWORKS_DM_SOWMYAJYOTHI.pdfNEURALNETWORKS_DM_SOWMYAJYOTHI.pdf
NEURALNETWORKS_DM_SOWMYAJYOTHI.pdf
 
Neural networks are parallel computing devices.docx.pdf
Neural networks are parallel computing devices.docx.pdfNeural networks are parallel computing devices.docx.pdf
Neural networks are parallel computing devices.docx.pdf
 
Artificial Neural Network in Medical Diagnosis
Artificial Neural Network in Medical DiagnosisArtificial Neural Network in Medical Diagnosis
Artificial Neural Network in Medical Diagnosis
 
lecture11_Artificial neural networks.ppt
lecture11_Artificial neural networks.pptlecture11_Artificial neural networks.ppt
lecture11_Artificial neural networks.ppt
 
Neural Network
Neural NetworkNeural Network
Neural Network
 
Artificial neural network by arpit_sharma
Artificial neural network by arpit_sharmaArtificial neural network by arpit_sharma
Artificial neural network by arpit_sharma
 
Jack
JackJack
Jack
 
Neural-Networks.ppt
Neural-Networks.pptNeural-Networks.ppt
Neural-Networks.ppt
 
Neural network
Neural networkNeural network
Neural network
 

Recently uploaded

Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
Joaquim Jorge
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
panagenda
 

Recently uploaded (20)

Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, AdobeApidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
Apidays New York 2024 - Scaling API-first by Ian Reasor and Radu Cotescu, Adobe
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivity
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Deploy with confidence: VMware Cloud Foundation 5.1 on next gen Dell PowerEdg...
Deploy with confidence: VMware Cloud Foundation 5.1 on next gen Dell PowerEdg...Deploy with confidence: VMware Cloud Foundation 5.1 on next gen Dell PowerEdg...
Deploy with confidence: VMware Cloud Foundation 5.1 on next gen Dell PowerEdg...
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
GenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdfGenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdf
 
HTML Injection Attacks: Impact and Mitigation Strategies
HTML Injection Attacks: Impact and Mitigation StrategiesHTML Injection Attacks: Impact and Mitigation Strategies
HTML Injection Attacks: Impact and Mitigation Strategies
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 

ANN-lecture9

  • 1. 5/7/2012 WHAT ARE NEURAL NETWORKS?  Neural networks are parallel information processing systems with their architecture inspired by the structure and functioning of the brain  A neural network have the ability to learn and generalize.  Neural networks can be trained to make classifications and predictions based on historical data  Neural nets are included in many data mining products 1
  • 2. 5/7/2012  Very popular and effective techniques  Artificial Neural Network goes by many names, such as connectionism, parallel distributed processing, neuro-computing and natural intelligent systems.  It is abbreviated by ANN  ANN has a strong similarity to the biological brain.  It is composed of interconnected elements called neurons. 3 The Biological Neuron  Biological neuron receives inputs from other sources, combines them in some way, performs a generally nonlinear operation on the result, and then output the final result. send Inputs Process the inputs (eyes, ears) Turn the processed inputs into outputs The electrochemical contact between neurons 4 2
  • 3. 5/7/2012 The Artificial Neuron 5 Network Layers  The neurons are grouped into layers.  The input layer.  The output layer.  Hidden layers between these two layers. 6 3
  • 4. 5/7/2012 Neural Network Architecture  In a Neural Network, neurons are grouped into layers.  The neurons in each layer are the same type.  There are different types of Layers.  The Input layer, consists of neurons that receive input from the external environment.  The Output layer, consists of neurons that communicate to the user or external environment.  The Hidden layer, consists of neurons that ONLY communicate with other layers of the network.  Now that we have a model for an artificial neuron, we can imagine connecting many of then together to form an Artificial Neural Network: Output layer Hidden layer Input layer 4
  • 5. 5/7/2012 Based on the following assumptions: 1. Information processing occurs at many simple processing elements called neurons. 2. Signals are passed between neurons over interconnection links. 3. Each interconnection link has an associated weight. 4. Each neuron applies an activation function to determine its output signal. A Neuron x0 - mk w0 x1 w1  f output y xn wn Input weight weighted Activation vector x vector w sum function  The n-dimensional input vector x is mapped into variable y by means of the scalar product and a nonlinear function mapping 5
  • 6. 5/7/2012 What is a neuron?  a (biological) neuron is a node that has many inputs and one output  f  inputs come from other neurons  the inputs are weighted  weights can be both positive and negative  inputs are summed at the node to produce an activation value The Artificial Neuron Model  In order to simulate neurons on a computer, we need a mathematical model of this node  Node j ( a neuron) has n inputs xi , i = 1 to n  Each input (connection) is associated with a weight wij  The neuron includes a bias denoted by bj  The bias has the effect of increasing or deceasing the net input 12 6
  • 7. 5/7/2012  The net input to node j (net result) ,uj , is the sum of the products of the connection inputs and their weights: n u j   w ijxi  b j i 1 Where uj is called the “activation of the neuron”.  The output of node j is determined by applying a non- linear transfer function g to the net input: x j  g(u j )  called the “transfer function”.  A common choice for the transfer function is the sigmoid: 1 g(u j )  u j 1 e  The sigmoid has similar non-linear properties to the transfer function of real neurons:  It accepts inputs varying from – ∞ to ∞  bounded below by 0  bounded above by 1 7
  • 8. 5/7/2012 An Example of multilayer feed forward neural network 8
  • 9. 5/7/2012 Architecture of ANN  Feed-Forward Networks The signals travel one way from input to output  Feed-Back Networks The signals travel as loops in the network, the output of the network is connected to the input. 17 Learning ANN  The purpose of the learning function is to modify the variable connection weights causes the network to learn the solution to a problem according to some neural based algorithm.  There are tow types of learning  Supervised (Reinforcement) learning.  Unsupervised learning. 18 9
  • 10. 5/7/2012 Supervised learning  Means their exist an external help or a teacher.  The teacher may be a training set of data.  The target is to minimize the error between the desired and the actual output.  The process take place as follows: 19  Presenting input and output data to the network, this data is often referred to as the “training set”.  The network processes the inputs and compares its resulting outputs against the desired outputs.  Errors are then propagated back through the system, causing the system to adjust the weights, which are usually randomly set to begin with.  This process occurs over and over as a closer match between the desired and the predicted output. 20 10
  • 11. 5/7/2012  When no further learning is necessary, the weights are typically frozen for the application.  There are many algorithms used to implement the adaptive feedback required to adjust the weights during training. The most common technique is called “Back-Propagation”. Let us suppose that a sufficiently large set of examples (training set) is available. Supervised learning: – The network answer to each input pattern is directly compared with the desired answer and a feedback is given to the network to correct possible errors 11
  • 12. 5/7/2012 Back Propagation (BP)  It is an improving performance method in training of multilayered feed forward neural networks.  Are used to adjust the weights and biases of networks to minimize the sum squared error of the network which is given by 1 SSE  2  (xt  xt )2 ˆ  where xt and xt^ are the desired and predicted output of the tth output node. 23 BP Network – Supervised Training  Desired output of the training examples  Error = difference between actual & desired output  Change weight relative to error size  Calculate output layer error , then propagate back to previous layer  Hidden weights updated  Improved performance 24 12
  • 13. 5/7/2012  Probably the most common type of ANN used today is a multilayer feed forward network trained using back-propagation (BP)  Often called a Multilayer Perceptron (MLP) Unsupervised learning  The training set consists of input training patterns but not with desired outputs.  This method often referred to as self-organization.  Therefore, the network is trained without benefit of any teacher. 26 13
  • 14. 5/7/2012  Applications in Clustering and reducing dimensionality  Learning may be very slow  No help from the outside  No training data, no information available on the desired output  Learning by doing  Used to pick out structure in the input:  Clustering  Compression 27 Choosing the network size  It seems better to start with a small number of neurons, because:  learning is faster.  it is often enough.  it avoids over-fitting problems.  If the number of neurons are too much we will get an over fit.  In principle, one hidden layer is sufficient to solve any problem. In practice, it may happen that two hidden layers with a small number of neurons may work better (and/or learn faster) than a network with a single layer. 28 14
  • 15. 5/7/2012 • Too few nodes: Don’t fit the curve very well • Too many nodes: Over parameterization • May fit noise • makes network more difficult to train The Learning Rate  The learning rate c, which determines by how much we change the weights w at each step.  If c is too small, the algorithm will take a long time to converge. Sum-Square Error Error Surface Epoch 30 15
  • 16. 5/7/2012  If c is too large, the network may not be able to make the fine discriminations possible with a system that learns more slowly. The algorithm diverges. Sum-Square Error Error Surface Epoch 31 Multi-Layer Perceptron Output vector Errj  O j (1  O j ) Errk w jk Output nodes k  j   j  (l) Errj wij  wij  (l ) Errj Oi Hidden nodes Errj  O j (1  O j )(T j  O j ) wij 1 Oj  I 1 e j Input nodes I j   wij Oi   j i Input vector: xi 16
  • 17. 5/7/2012 Applications of ANNs  Prediction – weather, stocks, disease, Predicting financial time series  Classification – financial risk assessment, image processing  Data Association – Text Recognition  Data Conceptualization – Customer purchasing habits  Filtering – Normalizing telephone signals  Optimization 33  Diagnosing medical conditions  Identifying clusters in customer databases  Identifying fraudulent credit card transactions  Hand-written character recognition  and many more…. 17
  • 18. 5/7/2012  Advantages  Adapt to unknown situations  Autonomous learning and generalization  Their most important advantage is in solving problems that are too complex for conventional technologies, problems that do not have an algorithmic solution or for which an algorithmic solution is too complex to be found.  Disadvantages  Not exact  Large complexity of the network structure 35 Using a neural network for prediction  Identify input and outputs  Preprocess inputs - often scale to the range [0,1]  Choose an ANN architecture  Train the ANN with a representative set of training examples (usually using BP)  Test the ANN with another set of known examples  often the known data set is divided in to training and test sets. Cross-validation is a more rigorous validation procedure.  Apply the model to unknown input data 18