lec# 1- introduction to ann

Upload: samer-kamel

Post on 06-Apr-2018

227 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 Lec# 1- Introduction to ANN

    1/8

    Inas A. YassineAssisstant Professor,

    Systems and Biomedical Engineering Department, CaroUniversity

    [email protected]

  • 8/3/2019 Lec# 1- Introduction to ANN

    2/8

    Course Syllabus Course Objective

    Describe the relation between human brain and simple ANNmodels.

    Explain the most common architectures and learningalgorithms for Multi-Layer Perceptrons, Radial-BasisFunction Networks, Committee Machines.

    Discuss the main factors involved in achieving good learningand generalization performance in neural network systems.

    Identify the main implementation issues for common neuralnetwork systems.

    Evaluate the practical considerations in applying neuralnetworks to real classification and regression problems.

  • 8/3/2019 Lec# 1- Introduction to ANN

    3/8

    Course Syllabus 'Cont.' Reference Book

    Neural Networks: A comprehensive Foundation, SimonHayken, Second Edition, 1999.

    Grading Policy

    Section, Exercises and projects (40 points)

    Mid term Exam (20 points)

    Term Exam (90 points)

  • 8/3/2019 Lec# 1- Introduction to ANN

    4/8

    What is Neural Network (NN) NN: are networks of neurons, for example, as found in real

    (i.e. biological) brains.

    Artificial Neurons are crude approximations of theneurons found in brains. They may be physical devices, orpurely mathematical constructs.

  • 8/3/2019 Lec# 1- Introduction to ANN

    5/8

    Artificial Neural Network (ANN) (ANNs) : are networks of Artificial Neurons, and hence

    constitute crude approximations to parts of real brains. Practically speaking, an ANN is

    just a parallel computationalsystem consisting of many simpleprocessing elements connectedtogether in a specific way in order

    to perform a particular task. Never to lose sight of how crude the approximations

    are, and how over-simplified our ANNs are comparedto real brains.

    http://upload.wikimedia.org/wikipedia/commons/e/e4/Artificial_neural_network.svghttp://upload.wikimedia.org/wikipedia/commons/e/e4/Artificial_neural_network.svghttp://upload.wikimedia.org/wikipedia/commons/e/e4/Artificial_neural_network.svghttp://upload.wikimedia.org/wikipedia/commons/e/e4/Artificial_neural_network.svg
  • 8/3/2019 Lec# 1- Introduction to ANN

    6/8

    Importance of ANNs powerful computational devices (Turing equivalent,

    universal computers).

    Massive parallelism makes them very efficient.

    can learn and generalize from training data.

    fault tolerant this is equivalent to the gracefuldegradation found in biological systems.

    very noise tolerant so they can cope with situations wherenormal symbolic systems would have difficulty.

    can do anything a symbolic/logic system can do, and more.(In practice, getting them to do it can be rather difficult)

  • 8/3/2019 Lec# 1- Introduction to ANN

    7/8

    Uses of ANNsthere are two basic goals for neural network research:

    Brain modelling : The scientific goal of building models of how realbrains work.

    This can potentially help us understand the nature of human intelligence,formulate better teaching strategies, or better remedial actions for braindamaged patients.

    Artificial System Building : The engineering goal of building efficientsystems for real world applications. This may make machines morepowerful, relieve humans of tedious tasks, and may even improve upon

    human performance. These should not be thought of as competing goals. We often use exactly the

    same networks and techniques for both. Frequently progress is made when thetwo approaches are allowed to feed into each other. There are fundamentaldifferences though, e.g. the need for biological plausibility in brain modelling,and the need for computational efficiency in artificial system building.

  • 8/3/2019 Lec# 1- Introduction to ANN

    8/8

    Learning Algorithms There are many forms of neural networks. Most operate by

    passing neural activations through a network ofconnected neurons.

    One of the most powerful features of neural networks istheir ability to learn andgeneralize from a set of trainingdata. They adapt the strengths/weights of the connectionsbetween neurons so that the final output activations arecorrect.

    There are three broad types of learning: . Supervised Learning (i.e. learning with a teacher)

    Reinforcement learning (i.e. learning with limited feedback)

    Unsupervised learning (i.e. learning with no help)