MELJUN CORTES IBM SPSS Neural Networks

Embed Size (px)

Citation preview

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    1/18

    Neural Networks

    This contains my personal notes only

    thus, this is not complete. Most of the

    contents were taken from the training

    manual of IBM SPSS Modeler. Please

    refer to the training manual for acomplete discussion.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    2/18

    Neural Networks

    It attempts to solve problems using methods that is

    similar on how the brains operate.

    Think of how a parent teaches a child how to read.

    Pattern of letterspresented to the

    child

    The child makes an attempt.

    If the child is correct, then she is rewarded. Thenext time she sees the same combination of letters

    she is likely to remember the correct response.

    If the child is incorrect, then she is told the

    correct response and tries to adjust her

    response based on this feedback.

    It starts with inputs. As the child receives the inputs, her

    brains work and then produce outputs.

    Neural networks work the same way!

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    3/18

    Neural Networks (NN)

    Attempts to solve problems using

    methods modeled on how the brain

    operates.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    4/18

    The input layer contains thefields used to predict theoutcome.

    The output layer contains theoutput field (the target ofprediction)

    The input and output fields canbe numeric or symbolic. Thesymbolic fields are transformedinto numeric (binary setcoding) before processing bythe network.

    The hidden layer contains anumber of neurons at whichoutputs from previous layercombine. A network can haveany number of hidden layers

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    5/18

    Neural Networks look like below:

    Neural network is learning the relationship between data and results

    itis said to be training.

    Once fully trained, the network can be given new unseen data and can

    make a decision/prediction based upon its experience.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    6/18

    Neural network consists of a number of

    processing elements (called neurons) thatare arranged in layers. Each neuron is

    linked to every neuron in the previous

    layer by connections that have strengthsor weights attached to them. The learning

    algorithm controls the adaptation of these

    weights to the data; this gives the system

    the capability to learn by example and

    generalize for new situations.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    7/18

    Main Consideration in building a network Locate

    the Global Solution!

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    8/18

    Different types of Neural Networks

    Multi-Layer Perceptron (MLP)

    Radial Basis Function Network (RBFN)

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    9/18

    Multi-Layer Perceptron (MLP)

    MLP networks consist of an input layer, an output layerand one or more hidden layers. The hidden layer isrequired to perform non-linear mappings.

    Each hidden layer neuron receives an input based on a

    weighted combination of the outputs of the neurons inthe previous layer.

    The neurons within the final hidden layer are, in turn,combined to produce an output. This predicted value isthen compared to the correct output and the difference

    between the two values (the error) is fed back into thenetwork, which in turn is updated. The feeding of theerror back through the network is referred to as back-propagation.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    10/18

    Child learning the difference

    between apple and orange: The child may decide in making a decision that the most

    useful factors are the shape, color, and size of the fruit these are the inputs.

    When shown the first example of a fruit she may look at

    the fruit and decide that it is round, red in color and of aparticular size.

    Not knowing of what an apple or an orange looks like,the child may decide to place equal importance on eachof such factors the importance is what a network refers

    to as weights.

    At this stage, the child is most likely to randomly chooseeither an apple or an orange for her prediction.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    11/18

    On being told the correct response, the child will

    increase or decrease the relative importance of

    each of the factors to improve her decision

    (reduce the error). In a similar fashion a MLP network begins with

    random weights placed on each of the inputs.

    On being told the correct response, the network

    adjusts these internal weights. In time, the childand the network will hopefully make correct

    predictions.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    12/18

    MLP fits a non-linear curve

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    13/18

    The advantages of using a MLP are:

    It is effective on a wide range of problems It is capable of generalizing well

    If the data are not clustered in terms of their inputfields, it will classify examples in the extreme regions

    It is currently the most commonly used type ofnetwork and there is much literature discussing itsapplications

    The disadvantages of using a MLP are: It can take a great deal of time to train

    It does not guarantee finding the best global solution

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    14/18

    MLP learning algorithms

    Quick

    Dynamic

    Multiple Prune

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    15/18

    Radial Basis Function Network (RBFN)

    It uses the k-means clusteringalgorithm to determine

    the number and location of the centers in the input space

    The RBF can be thought of

    performing a type of clustering

    within the input space, encircling

    individual clusters of data by a

    number of basis functions. If a data

    point falls within the region of

    activation of a particular basis

    function, then the neuron

    corresponding to that basis functionresponds most strongly.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    16/18

    The advantages of using a RBF network are:

    It is quicker to train than a MLP

    It can model data that are clustered within the input

    space.The disadvantages of using a RBF network are:

    It is difficult to determine the optimal position of the

    function centers

    The resulting network often has a poor ability torepresent the global properties of the data.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    17/18

    Which Method to Use?

    When building neural networks it is

    sensible to try both algorithms and

    either choose the one with the bestoverall performance, or, use both

    models to gain a majority prediction.

  • 7/30/2019 MELJUN CORTES IBM SPSS Neural Networks

    18/18

    Neural Network Hands-on

    Data set: Churn.txt