MTR607 Chapter 0

Embed Size (px)

Citation preview

  • 7/28/2019 MTR607 Chapter 0

    1/74

    Learning Algorithm and

    Neural Networks

  • 7/28/2019 MTR607 Chapter 0

    2/74

    MTR 607

    Textbook:Simon Haykin, Neural Networks A Comprehensive

    Foundation, 2nd Ed., 1999

    Lecturer:Dr. Alaa Sagheer

    Place:Seminar Room, E-JUST

    Grading: Class participation (10%),

    Assignments and reports (20%),

    Midterm test (30%),

    Final exam (40%)

    2

  • 7/28/2019 MTR607 Chapter 0

    3/74

    Course OverviewIntroduction to Artificial Neural Networks,

    Artificial and human neurons (Biological Inspiration)

    The learning process,

    Supervised and unsupervised learning,

    Reinforcement learning,

    Applications Development and Portfolio

    The McCulloch-Pitts Model of Neuron,

    A simple network layers, Multilayer networks

    Perceptron,

    Back propagation algorithm,

    Recurrent networks,Associative memory,

    Self Organizing maps,

    Support Vector Machine and PCA,

    Applications to speech, vision and control problems.

    3MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    4/74

    ANNs ResourcesMain text books:

    Neural Networks: A Comprehensive Foundation

    , S. Haykin(very good -

    theoretical)

    Pattern Recognition with Neural Networks, C. Bishop (very good-more accessibleNeural Network Design by Hagan, Demuth and Beale (introductory)

    Books emphasizing the practical aspects:

    Neural Smithing, Reeds and MarksPractical Neural Network Recipees in C++ T. Masters

    Seminal Paper:

    Parallel Distributed Processing Rumelhart and McClelland et al.

    Other:

    Neural and Adaptive Systems, J. Principe, N. Euliano, C. Lefebvre

    4MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    5/74

    Review Articles:R. P. Lippman, An introduction to Computing with Neural Nets IEEE ASPMagazine, 4-22, April 1987.

    T. Kohonen, An Introduction to Neural Computing, Neural Networks, 1, 3-16, 1988.A. K. Jain, J. Mao, K. Mohuiddin, Artificial Neural Networks: A Tutorial IEEEComputer, March 1996 p. 31-44.

    ANNs Resources

    5MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    6/74

    Course OverviewIntroduction to Artificial Neural Networks,

    Artificial and human neurons (Biological Inspiration)The learning process,

    Supervised and unsupervised learning,

    Reinforcement learning,

    Applications Development and Portfolio

    The McCulloch-Pitts Model of Neuron,A simple network layers, Multilayer networks

    Perceptron,

    Back propagation algorithm,

    Recurrent networks,

    Associative memory,Self Organizing maps,

    Support Vector Machine and PCA,

    Applications to speech, vision and control problems.

    6MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    7/74

    Part I:

    1. Artificial Neural Networks

    2. Artificial and human neurons (Biological Inspiration)

    3. Tasks & Applications of ANNs

    Part II:

    1. Learning in Biological Systems

    2. Learning with Artificial Neural Networks

    Introduction to Artificial Neural Networks

    7MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    8/74

    ANNs vs. ComputersDigital Computers

    Analyze the problem to be solved

    Deductive Reasoning. We applyknown rules to input data to produce

    output.

    Computation is centralized,synchronous, and serial.

    Not fault tolerant. One transistor goesand it no longer works.

    Static connectivity.

    Applicable if well defined rules withprecise input data.

    8MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Artificial Neural Networks

    No requirements of an explicit

    description of the problem.

    Inductive Reasoning. Given input and

    output data (training examples), we

    construct the rules.

    Computation is collective,asynchronous, and parallel.

    Fault tolerant and sharing of

    responsibilities.

    Dynamic connectivity.

    Applicable if rules are unknown or

    complicated, or if data are noisy or

    partial.

  • 7/28/2019 MTR607 Chapter 0

    9/74

    What is ANN?

    9MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    ANN is a branch of "Artificial Intelligence". It is a system modeled based on the human brain.ANN goes by many names, such as connectionism, parallel distributed processing, neuro-computing, machine learning algorithms, and finally, artificial neural networks.

    Developing ANNs date back to the early 1940s. It experienced a wide popularity in the late

    1980s. This was a result of the discovery of new techniques and developments in PCs.

    Some ANNs are models of biological neural networks and some are not.

    ANN is a processing device (An algorithm or Actual hardware) whose design was motivated by

    the design and functioning of human brain.

    Inside ANN:

    ANNs design is what distinguishes neural networks from other mathematical techniquesANN is a network of many simple processors ("units or neurons), each unit has a smallamount of local memory.

    The units are connected by unidirectional communication channels ("connections"), whichcarry numeric (as opposed to symbolic) data.

    The units operate only on their local data and on the inputs they receive via the connections.

    Artificial Neural Networks (1)

  • 7/28/2019 MTR607 Chapter 0

    10/74

    ANNs Operation

    ANNs normally have great potential for parallelism (multiprocessor-friendly architecture),

    since the computations of the units are independent of each other. Same like biological neuralnetworks.

    Most neural networks have some kind of "training" rule whereby the weights of connections areadjusted on the basis of presented patterns.

    In other words, neural networks "learn" from examples, just like childrenand exhibit somestructural capability for generalization.

    10MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Artificial Neural Networks (2)

  • 7/28/2019 MTR607 Chapter 0

    11/74

    ANNs are a powerful technique (Black Box) to solve many real world

    problems. They have the ability to learn from experience in order toimprove their performance and to adapt themselves to changes in the

    environment.

    In addition, they are able to deal with incomplete information or

    noisy data and can be very effective especially in situations where it is

    not possible to define the rules or steps that lead to the solution of aproblem.

    Once trained, the ANN is able to recognize similarities when

    presented with a new input pattern, resulting in a predicted output

    pattern.

    11MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Artificial Neural Networks (3)

  • 7/28/2019 MTR607 Chapter 0

    12/74

    What can a ANN do?

    Compute a known function

    Approximate an unknown function

    Pattern Recognition

    Signal Processing

    .Learn to do any of the above

  • 7/28/2019 MTR607 Chapter 0

    13/74

    Part I:

    1. Artificial Neural Networks (ANNs)

    2. Artificial and human neurons (Biological Inspiration)

    3. Tasks & Applications of ANNs

    Part II:

    1. Learning in Biological Systems

    2. Learning with Artificial Neural Networks

    Introduction to Artificial Neural Networks

    13MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    14/74

    Biological Inspiration

    Animals are able to react adaptively to changes in their external

    and internal environment, and they use their nervous system toperform these behaviours.

    An appropriate model/simulation of the nervous system shouldbe able to produce similar responses and behaviours in artificialsystems.

    The nervous system is build by relatively simple units, theneurons, so copying their behaviour and functionality should bethe solution!

    Biological Neural Networks (BNN) are much more

    complicated in their elementary structures than the

    mathematical models we use for ANNs

    14MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    15/74

    ANN as a model of brain-

    like Computer

    Brain

    The human brain is still not well

    understood and indeed its behavior

    is very complex!

    There are about 10-11 billion

    neurons in the human cortex each

    connected to , on average, 10000others. In total 60 trillion synapses

    of connections.

    The brain is a highly complex,

    nonlinear and parallel computer

    (information-processing system)

    ANN as a Brain-Like Computer

    15MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    An artificial neural network (ANN) is

    a massively parallel distributed processor

    that has a natural propensity for storingexperimental knowledge and making it

    available for use. It means that:

    Knowledge is acquired by the network

    through a learning (training) process;

    The strength of the interconnections

    between neurons is implemented by

    means of the synaptic weights used to

    store the knowledge.

    The learning process is a procedure of theadapting the weights with a learning

    algorithm in order to capture the knowledge.

    On more mathematically, the aim of the

    learning process is to map a given relation

    between inputs and output of the network.

  • 7/28/2019 MTR607 Chapter 0

    16/74

    Massive parallelism

    Brain computer as an information

    or signal processing system, is

    composed of a large number of a

    simple processing elements, calledneurons. These neurons are

    interconnected by numerous direct

    links, which are called connection,

    and cooperate which other to

    perform a parallel distributed

    processing (PDP) in order to soft a

    desired computation tasks.

    Connectionism

    Brain computer is a highly

    interconnected neurons system in

    such a way that the state of one

    neuron affects the potential of thelarge number of other neurons

    which are connected according to

    weights or strength. The key idea

    of such principle isthe functional

    capacity of biological neural nets

    deters mostly not so of a single

    neuron but of its connections

    Associative

    distributed memory

    Storage of information in a brain is

    supposed to be concentrated insynaptic connections of brain

    neural network, or more precisely,

    in the pattern of these connections

    and strengths (weights) of the

    synaptic connections.

    A process of pattern

    recognition and pattern

    manipulation is based on:

    How our brainmanipulates

    with patterns ?

    Principles of Brain Processing

    16MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    17/74

    17MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Biological

    Neuron

    - The simple

    arithmeticcomputing

    element

    Biological Neuron

  • 7/28/2019 MTR607 Chapter 0

    18/74

    Cell structures

    Cell body

    DendritesAxon

    Synaptic terminals

    Biological Neuron (2)

    18MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    19/74

    synapses

    axon dendrites

    The information transmission happens at the synapses, i.e

    Synaptic connection strengths among neurons are used to storethe acquired knowledge.

    In a biological system, learning involves adjustments to the

    synaptic connections between neurons

    19MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Biological Neurons (3)

  • 7/28/2019 MTR607 Chapter 0

    20/74

    Biological Neurons (4)

    20MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    1. Soma or body cell - is a large, round

    central body in which almost all the

    logical functions of the neuron are

    realized (i.e. the processing unit).

    2. The axon (output), is a nerve fibre

    attached to the soma which can serve

    as a final output channel of the

    neuron. An axon is usually highly

    branched.

    3. The dendrites (inputs)- represent a

    highly branching tree of fibers. These

    long irregularly shaped nerve fibers

    (processes) are attached to the soma

    carrying electrical signals to the cell

    4. Synapses are the point of contact

    between the axon of one cell and the

    dendrite of another, regulating a

    chemical connection whose strength

    affects the input to the cell.

    The schematic

    model of a

    biological neuron

    Synapses

    Dendrites

    Soma

    AxonDendrite

    from

    other

    Axon from

    other

    neuron

  • 7/28/2019 MTR607 Chapter 0

    21/74

    Properties of ANNs

    Learning from examples

    labeled or unlabeled

    Adaptivity

    changing the connection strengths to learn things

    Non-linearity

    the non-linear activation functions are essential

    Fault tolerance

    if one of the neurons or connections is damaged, the

    whole network still works quite well

  • 7/28/2019 MTR607 Chapter 0

    22/74

    Part I:

    1. Artificial Neural Networks (ANNs)

    2. Artificial and human neurons (Biological Inspiration)

    3. Tasks & Applications of ANNs

    Part II:

    1. Learning in Biological Systems

    2. Learning with Artificial Neural Networks

    Introduction to Artificial Neural Networks

    22MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    23/74

    Applications of ANNsClassification

    In marketing: consumer spending pattern classification

    In defence: radar and sonar image classification

    In agriculture & fishing: fruit, fish and catch grading

    In medicine: ultrasound and electrocardiogram image classification, EEGs, medical diagnosis

    Recognition and Identification

    In general computing and telecommunications: speech, vision and handwriting recognition

    In finance: signature verification and bank note verificationAssessment

    In engineering: product inspection monitoring and control

    In defence: target tracking

    In security: motion detection, surveillance image analysis and fingerprint matching

    Forecasting and Prediction

    In finance: foreign exchange rate and stock market forecasting

    In agriculture: crop yield forecasting , Deciding the category of potential food items

    (e.g., edible or non-edible)

    In marketing: sales forecasting

    In meteorology: weather prediction

    MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer 23

  • 7/28/2019 MTR607 Chapter 0

    24/74

    Who are the Men of ANNs?!Computer scientists want to find out about the properties of non-symbolicinformation processing with neural nets and about learning systems ingeneral.

    Statisticians use neural nets as flexible, nonlinear regression andclassification models.

    Engineers of many kinds exploit the capabilities of neural networks in manyareas, such as signal processing and automatic control.

    Cognitive scientists view neural networks as a possible apparatus to describemodels of thinking and consciousness (High-level brain function).

    Neuro-physiologists use neural networks to describe and explore medium-level brain function (e.g. memory, sensory system, motorics).

    Physicists use neural networks to model phenomena in statistical mechanics

    and for a lot of other tasks.Biologists use Neural Networks to interpret nucleotide sequences.

    Philosophers and some other people may also be interested in NeuralNetworks for various reasons

    24MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    25/74

    Operation of Biological Neuron

    The spikes travelling along the axon of the pre-synaptic neurontrigger the release of neurotransmitter substances at thesynapse.

    The neurotransmitters cause excitation or inhibition in thedendrite of the post-synaptic neuron.

    The integration of the excitatory and inhibitory signals mayproduce spikes in the post-synaptic neuron.

    The contribution of the signals depends on the strength of thesynaptic connection.

    25MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Excitation means positive product between the incoming

    spike rate and the corresponding synaptic weight;

    Inhibition means negative product between the incomingspike rate and the corresponding synaptic weight;

  • 7/28/2019 MTR607 Chapter 0

    26/74

    Inputs

    Output

    An artificial neural network is composed of many

    artificial neurons that are linked together accordingto a specific network architecture. The objective of

    the neural network is to transform the inputs into

    meaningful outputs.

    ANN Architecture

    26MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    27/74

    Neurons are arranged in layers. Neurons work by processing information. They

    receive and provide information in form of spikes.

    27MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    The artificial neuron receives one or more inputs (representing the one or more

    dendrites),

    At each neuron, every input has an associated weight which modifies the

    strength of each input and sums them together,

    The sum of each neuron is passed through a function known as an activation

    function ortransfer function in order to produce an output (representing a

    biological neuron's axon)

    ANN Architecture (2)

    Inputs Output

    http://en.wikipedia.org/wiki/Dendritehttp://en.wikipedia.org/wiki/Activation_functionhttp://en.wikipedia.org/wiki/Activation_functionhttp://en.wikipedia.org/wiki/Transfer_functionhttp://en.wikipedia.org/wiki/Axonhttp://en.wikipedia.org/wiki/Axonhttp://en.wikipedia.org/wiki/Transfer_functionhttp://en.wikipedia.org/wiki/Activation_functionhttp://en.wikipedia.org/wiki/Activation_functionhttp://en.wikipedia.org/wiki/Dendrite
  • 7/28/2019 MTR607 Chapter 0

    28/74

    Input

    s

    Outputw2

    w1

    w3

    wn

    wn-1

    . . .

    x1

    x2

    x3

    xn-1

    xn

    y)(;

    1

    zHyxwzn

    i

    ii

    Each neuron takes one or more inputs and produces an output. At eachneuron, every input has an associated weight which modifies the strength of

    each input. The neuron simply adds together all the inputs and calculates an

    output to be passed on.

    28MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    ANN Architecture (3)

    f A

  • 7/28/2019 MTR607 Chapter 0

    29/74

    Models of A Neuron

    29MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    M d l f A N (2)

  • 7/28/2019 MTR607 Chapter 0

    30/74

    Axon

    Terminal Branches

    of AxonDendrites

    S

    x1

    x2

    w1

    w2

    wn

    xn

    x3 w3

    Models of A Neuron (2)

    30MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    M d l f A N (3)

  • 7/28/2019 MTR607 Chapter 0

    31/74

    1. A set of synapses, or connection link: each of whichis characterized by a weight or strength of its own wkj.

    Specifically, a signal xj at the input synapse j connected toneuron k is multiplied by the synaptic wkj

    2. An adder: For summing the input signals, weighted byrespective synaptic strengths of the neuron in a linear

    operation.

    3. Activation function: For limiting of the amplitude of theoutput of the neuron to limited range. The activation function

    is referred to as a Squashing (i.e. limiting) function {interval

    [0,1], or, alternatively [-1,1]}

    Models of A Neuron (3)

    Three elements:

    31MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    32/74

    BiasThe bias has the effect of increasing or lowering the net input of

    the activation function depending on whether it is +/-

    yk= (vk) = (uk+ bk) = (S wkjxj + bk)An artificial neuron:

    - computes the weighted sum of its input (called its net input)- adds its bias (the effect of applying affine transformation to the output vk)

    - passes this value through an activation function

    We say that the neuron fires (i.e. becomes active) if

    its outputs is above zero.This extra free variable (bias) makes the neuron more

    powerful.

    32MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    33/74

    Activation Function (vk)It defines the output of the neuron given an input or set of inputs. A standard

    computer chip circuit can be seen as a digital network of activation functions

    that can be "ON" (1) or "OFF" (0), depending on input,

    The best activation function is the non-linear function. Linear functions are

    limited because the output is simply proportional to the input.

    33MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    Three basic types of activation function:1. Threshold function,

    2. Linear function,

    3. Sigmoid function.

    A ti ti f ti (2)

  • 7/28/2019 MTR607 Chapter 0

    34/74

    Threshold (Step) function

    The outputykof this activation function is binary, depending onwhether the input meets a specified threshold. The "signal" is sent,

    i.e. the output is set to one, if the activation meets the threshold.

    Activation functions (2)

    McColloch-Pitts ModelThreshold Logic Unit

    (TLU),since 194334MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    A ti ti f ti (3)

  • 7/28/2019 MTR607 Chapter 0

    35/74

    Piecewise Linear Function- The amplification factor inside the linear region of operation is assumed to beunity.

    - This form may be viewed as an approximation to a non linear amplifier

    Activation functions (3)

    35MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    A ti ti f ti (4)

  • 7/28/2019 MTR607 Chapter 0

    36/74

    Sigmoid function

    Where a is the slope parameter ofthe sigmoid function

    Activation functions (4)

    - A fairly simple non-linear function, such as the logistic function.- As the slop parameter approaches infinity the sigmoid function becomes a

    threshold function

    36MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

  • 7/28/2019 MTR607 Chapter 0

    37/74

    Artificial Neural Networks

    Early ANN Models:

    McCulloch-Pitts , Perceptron, ADALINE,

    Hopfield Network,

    Current Models:

    Multilayer feed forward networks (Multilayer

    perceptrons- Back propagation )

    Radial Basis Function networks

    Self Organizing Networks

    ...

    37MTR607: Learning Algorithms and Neural Networks Dr. Alaa Sagheer

    F db k

  • 7/28/2019 MTR607 Chapter 0

    38/74

    FeedbackFeedback is a dynamic system whenever occurs in

    almost every part of the nervous system,

    Feedback is giving one or more closed path for

    transmission of signals around the system,

    It plays important role in study of special class of

    neural networks known as Recurrent networks.

    F db k (2)

  • 7/28/2019 MTR607 Chapter 0

    39/74

    Feedback (2)

    The system is assumed to be linear and has a forward path (A)and a feedback path (B),

    The output of the forward channel determines its own output

    through the feedback channel.

    F db k (3)

  • 7/28/2019 MTR607 Chapter 0

    40/74

    Feedback (3)

    E.g. considerA is a fixed weight and B is a unit delay operatorz-1 .

    F db k (4)

  • 7/28/2019 MTR607 Chapter 0

    41/74

    Feedback (4)

    Then, we may expressyk(n) as an infinite weighted summation of

    present and past samples of the input signalxj(n).

    Therefore, feedback systems are controlled by weight.

    F db k (5)

  • 7/28/2019 MTR607 Chapter 0

    42/74

    Feedback (5)Feedback systems are controlled

    by weight.

    1. For positive weight, we have

    stable systems, i,e, convergent

    output y,

    2. For negative weight, we have,

    unstable systems, i.e divergent

    output y.. (Linear andExponential)

    N t k A hit t

  • 7/28/2019 MTR607 Chapter 0

    43/74

    Network ArchitecturesThree different classes of network architectures:

    1. Single-layer feed forward networks,

    2. Multilayer feed forward networks,

    3. Recurrent networks.

    Si l l f d f d t k

  • 7/28/2019 MTR607 Chapter 0

    44/74

    Single-layer feed forward network

    - Input layer of source nodes that projects directly

    onto an output layer of neurons.

    - Single-layer referring to the output layer ofcomputation nodes (neuron).

    M ltil f d f d t k

  • 7/28/2019 MTR607 Chapter 0

    45/74

    Multilayer feed forward networkIt contains one or more hidden

    layers (hidden neurons).Hidden refers to the part ofthe neural network is not seen

    directly from either input or

    output of the network .The function of hidden neuron

    is to intervene between input

    and output.

    By adding one or more hiddenlayers, the network is able to

    extract higher-order statistics

    from input

    R t N t k

  • 7/28/2019 MTR607 Chapter 0

    46/74

    Recurrent NetworksIt is different from feed

    forward neural network in thatit has at least one feedback

    loop.

    Recurrent network may consist

    of single layer of neuron witheach neuron feeding its output

    signal back to the inputs of all

    the other neurons.Note: There

    are no self-feedback.

    Feedback loops have a

    profound impact on learning

    and overall performance.

    H t D id N t k T l ?

  • 7/28/2019 MTR607 Chapter 0

    47/74

    Slide 47

    What transfer function should be used?

    How many inputs does the network need?

    How many hidden layers does the network need?

    How many hidden neurons per hidden layer?

    How many outputs should the network have?

    There is no standard methodology to determinate these values.

    Even there is some heuristic points, final values are

    determinate by a trial and error procedure.

    How to Decide on a Network Topology?

    K l d R t ti

  • 7/28/2019 MTR607 Chapter 0

    48/74

    Knowledge Representation

    The main characteristic of knowledge representation has two

    folds:

    1) What information is actually made explicit?

    2) How the information is physically encoded for subsequent use?

    Knowledgeis referred to the stored information or models used

    by a person or machine to interpret, predict and, appropriately,respond to the outside.

    A good solution depends on a good representation of

    knowledge

    K l d R t ti (2)

  • 7/28/2019 MTR607 Chapter 0

    49/74

    There are two kinds of Knowledge:1) The known world states, or facts, (prior knowledge),

    2) Observations (measurements) of the world, obtained by

    sensors to probe the environment.

    Knowledge Representation (2)

    These observations

    represent the pool of

    information, from

    which examples areused to train the NN

    K l d R t ti (3)

  • 7/28/2019 MTR607 Chapter 0

    50/74

    These Examples can be labeled or unlabeled

    In labeled examples

    Each example representing an input signal is paired with a

    corresponding desired response,

    Labeled examples may be expensive to collect, as they requireavailability of a teacher to provide a desired response foreach labeled example.

    Un labeled examples

    Unlabeled examples are usually abundant as there is no needfor supervision.

    Knowledge Representation (3)

    K l d R t ti (3)

  • 7/28/2019 MTR607 Chapter 0

    51/74

    Design of neural network may proceed as follow:An appropriate architecture for the neural network, with aninput layer consisting of source nodes equal in number to the

    pixels of an input image.

    The recognition performance of trained network is tested with

    data not seen before (testing).

    This phase of the network design called learning

    Knowledge Representation (3)

    R l f K l d R t ti

  • 7/28/2019 MTR607 Chapter 0

    52/74

    Roles of Knowledge RepresentationThere are four rules for knowledge representation:

    Rule 1:

    Similar inputs (i.e., patterns) drawn from similar

    classes should usually produce similar representation

    inside the network, and should therefore be classified as

    belonging to the same class.

    There are plethora (many) of measures for

    determining the similarity between inputs

    R l f K l d R t ti (2)

  • 7/28/2019 MTR607 Chapter 0

    53/74

    (1)

    Roles of Knowledge Representation (2)A commonly used measure of similarity is the Euclidian Distance

    Let xi denotes an m-by-1 vector

    R l f K l d R t ti (3)

  • 7/28/2019 MTR607 Chapter 0

    54/74

    Roles of Knowledge Representation (3)Another measure is the dot productorinner product com

    Given a pair of vectors xi andxj of the same dimension, their innerproduct will be (the projection of vectorx

    ionto vectorx

    j)

    Please note that:

    R l f K l d R t ti (4)

  • 7/28/2019 MTR607 Chapter 0

    55/74

    Using Eq.(1) to write

    Roles of Knowledge Representation (4)The smaller the Euclidean distance x

    i - x

    j(i.e. the more similar

    the vectorxi andxj are), the larger the inner product xiT

    xj will be.To formalize this relationship, we normalizethe vectors x

    i andx

    jto have a unit length, i.e.:

    The minimization of the Euclidean distance d(xi, xj ) corresponds

    to maximization of the inner product (xi, x

    j )..and, therefore, the

    similarity between the vectors xi and xj

    R l f K l d R t ti (5)

  • 7/28/2019 MTR607 Chapter 0

    56/74

    If the vectors xi and x

    jare stochastic (drown from different

    population of data)

    Roles of Knowledge Representation (5)

    Where C-1 is the inverse of the covariance

    matrix C. It is supposed that the

    covariance matrix is the same for both

    For a prescribed C, the smaller the distance d is themore similar the vectors xi and xj will be

    R l f K l d R t ti (6)

  • 7/28/2019 MTR607 Chapter 0

    57/74

    Rule 2:

    Item to be categorized as separate classes should be given widelydifferent representation in work.

    Rule 3:If a particular feature is important, then there should be large

    number of neurons involved in the representation of that item inthe network.

    Rule 4:Prior information and invariance should be built into the design of

    a neural network when ever they are available, so as to simplifythe network design by its not having to learn them.

    Roles of Knowledge Representation (6)

    Rule 4 is particularly important and highly desirable

    Roles of Kno ledge Representation (7)

  • 7/28/2019 MTR607 Chapter 0

    58/74

    1) Biological visual and auditory networks are very specialized,

    2) NN with SS has a smaller number of free parameters available for

    adjustment than other networks. Then, they need a small training dataset,

    learns faster and generalize better.

    3) Rate of information transmission through a specialized network is faster,

    4) Cost of building a specialized network is minimum, due to small size.

    Roles of Knowledge Representation (7)Rule 4 is particularly important and highly desirable

    because it results in an NN with a Specialized Structure (SS)

    How to build prior information into NN design?

  • 7/28/2019 MTR607 Chapter 0

    59/74

    How to build prior information into NN design?There are currently no well-defined rules for doing this; but wehave some procedure are known to yield useful rules. In particular,we may use a combination of two techniques:

    1. Restricting the network architecture (using local connections)

    2. Constraining the choice of synaptic weight (using the weightsharing)

    The latter tech is so

    important because it

    leads to reducing

    significantly freeparameters

    How to build invariance into NNs design?

  • 7/28/2019 MTR607 Chapter 0

    60/74

    How to build invariance into NNs design?

    There are three technique for rendering classifier-type NNs

    invariant to transformations:

    1. Invariance by structure.

    2. Invariance by training.

    3. Invariance by feature space

    Consider any of the following:

    1) When an object rotates, the perceived image, by observer, will change as well,

    2) The utterance of a spoken person may be soft or loud..slower or quicker,

    3) ..

    A classifier should be invariant to different transformation

    Or

    A class estimate represented by an output of the classifierMUST not be affected by transformations of the observed

    signal applied to the classifier input

  • 7/28/2019 MTR607 Chapter 0

    61/74

    Learning in

    Biological Systems

    Learning in Biological Systems

  • 7/28/2019 MTR607 Chapter 0

    62/74

    Learning in Biological Systems

    Learning approach based on modeling adaptation in

    biological neural systems

    Learning = learning by adaptation

    The young animal learns that the green fruits are sour,

    while the yellowish/reddish ones are sweet. The

    learning happens by adapting the fruit picking

    behaviour

    Learning in Biological Systems (2)

  • 7/28/2019 MTR607 Chapter 0

    63/74

    From experience: examples / training data

    Learning happens by changing of the synaptic

    strengths,

    Synapses change size and strength with experience (or

    examples or training data),

    Strength of connection between the neurons is stored

    as a weight-value for the specific connection,

    Learning the solution to a problem = changing the

    connection weights

    Learning in Biological Systems (2)

    Learning in Biological Systems (3)

  • 7/28/2019 MTR607 Chapter 0

    64/74

    Hebbian Learning

    When two connected neurons are firing at the same

    time, the strength of the synapse between them

    increases,

    Neurons that fire together, wire together

    Learning in Biological Systems (3)

    Learning in ANN

  • 7/28/2019 MTR607 Chapter 0

    65/74

    Learning in ANNWe may categorize the learning process through Neural

    Networks function as follows:

    1. Learning with a teacher,

    - Supervised Learning

    2. Learning without a teacher,

    - Unsupervised Learning

    - Reinforcement Learning

    Supervised Learning

  • 7/28/2019 MTR607 Chapter 0

    66/74

    In supervised learning, both the

    inputs and the outputs are

    provided. The network then

    processes the inputs and compares

    its resulting outputs against the

    desired outputs

    Errors are then calculated, causingthe system to adjust the weights

    which control the network. This

    process occurs over and over as the

    weights are continually improved.

    Supervised Learning

    Supervised learning process

    constitutes a closed-loop

    feedback system but unknown

    environment is outside the loop,

    Supervised Learning (2)

  • 7/28/2019 MTR607 Chapter 0

    67/74

    It is based on a labeled

    training set.

    The class of each piece

    of data in training set is

    known.

    Class labels are pre-

    determined and

    provided in the trainingphase.

    A

    B

    A

    BA

    B

    Class

    Class

    Class

    Class

    Class

    Class

    Supervised Learning (2)

    Understanding Supervised Learning

  • 7/28/2019 MTR607 Chapter 0

    68/74

    g p g

    A

    BA

    B A

    B

    Two Possible Solutions

  • 7/28/2019 MTR607 Chapter 0

    69/74

    Two Possible Solutions

    A

    B

    A

    B

    A

    B

    A

    B A

    B

    A

    B

    How to solve a given problem of supervised learning?

  • 7/28/2019 MTR607 Chapter 0

    70/74

    How to solve a given problem of supervised learning?Various steps have to be considered:

    1. Determine the type of training examples,

    2. Gather a training data set that satisfactory describe the given problem,

    3. After the training process we can test the performance of learned artificial

    neural network with the test (validation) data set,

    4. Test data set consist of data that has not been introduced to artificial

    neural network while learning.

    Reinforcement Learning

  • 7/28/2019 MTR607 Chapter 0

    71/74

    Reinforcement Learning

    The learning of inputoutputmapping is performed through

    continued interaction with the

    environment in order to minimize

    a scalar index of performance.

    Or

    A machine learning technique

    that sets parameters of an

    artificial neural network, where

    data is usually not given, but

    generated by interactions with the

    environment.

    Reinforcement Learning (2)

  • 7/28/2019 MTR607 Chapter 0

    72/74

    Reinforcement Learning (2)

    Reinforcement learning is built around critic that converts primary

    reinforcement signal received from the environment into a higher

    quality reinforcement signal

    Unsupervised Learning

  • 7/28/2019 MTR607 Chapter 0

    73/74

    U supe v sed e g

    No help from the outside,

    No information available on the desired output,

    Input: set of patterns P, from n-dimensional space S, but little /no information about their classification, evaluation, interestingfeatures, etc.

    It must learn these by itself!

    Learning by doingTasks: Used to pick out structure in the input

    Clustering - Group patterns based on similarity,

    Vector Quantization - Fully divide up S into a small set ofregions (defined by codebook vectors) that also helps cluster P,

    Feature Extraction - Reduce dimensionality of S by removingunimportant features (i.e. those that do not help in clustering P)

    Supervised vs. Unsupervised

  • 7/28/2019 MTR607 Chapter 0

    74/74

    Supervised vs. Unsupervised

    Task performed

    Classification

    Pattern Recognition

    NN model

    Preceptron,Feed-Forward NN

    Task performed

    Clustering, Pattern Recognition

    Feature Extraction, VQ

    NN Model

    Self Organizing Maps,ART