Artificial Neural Networking 1

download Artificial Neural Networking 1

of 18

Transcript of Artificial Neural Networking 1

  • 8/8/2019 Artificial Neural Networking 1

    1/18

    A Tribute To

    Mr.Frank Rosenblatt

    Father of Artificial Neuron Networking

  • 8/8/2019 Artificial Neural Networking 1

    2/18

    SEMINAR ONARTIFICIAL NEURAL NETWORKING

    By

    Arkaj yoti Bhattacharjee

    Under The Guidance Of :-

    Mr.Manas Ranjan Nayak

    Roll-10,Regd no-0505247099

  • 8/8/2019 Artificial Neural Networking 1

    3/18

    INTRODUCTION

    There is no known algorithm for predicting solventaccessibilit y or coordination number.

    Man y different approaches were tried, and most of them utilized the concept of neural networks .

    We shall discuss what these networks are, how dothey work, and how we use them for our cause.

  • 8/8/2019 Artificial Neural Networking 1

    4/18

    ARTIFICIAL NEURAL

    NETWORK Attempts to mimic the actions of the neuralnetworks of the human bod y

    Lets first look at how a biological neuralnetwork works A neuron is a single cell that conducts a

    chemicall y-based electronic signal At an y point in time a neuron is in either an

    excited or inhibited state

  • 8/8/2019 Artificial Neural Networking 1

    5/18

    S TRUCTURE OF A NEURON

    A series of connected neurons forms a pathwa y A series of excited neurons creates a strong

    pathwa y A biological neuron has multiple input tentacles

    called dendrites and one primar y output tentaclecalled an axon

    The gap between an axon and a dendrite is called asynapse

  • 8/8/2019 Artificial Neural Networking 1

    6/18

    NEURAL NETWORKINGIN A BIOLOGICAL CELL

  • 8/8/2019 Artificial Neural Networking 1

    7/18

    ARTIFICIAL NEURAL

    NETWORK S

    Each processing element in an artificial neuralnet is analogous to a biological neuron

    An element accepts a certain number of inputvalues and produces a single output value of either 0 or 1

    Associated with each input value is a numeric

    weight

  • 8/8/2019 Artificial Neural Networking 1

    8/18

    FEATURE S OF ANN

    NNs attempt to model the wa y the brain is structured:

    10 billion neurons that communicate via 60 trillionconnections (s ynapses).

    Parallel rather than sequential processing.

    NNs are composed of the following elements:

    Neuron (soma) Inputs (dendrites)

    Outputs of Neurons (axons)

    Weights (s ynapse)

  • 8/8/2019 Artificial Neural Networking 1

    9/18

    THE ACTIVITIE S WITHIN A

    PROCESS

    ING UNIT

  • 8/8/2019 Artificial Neural Networking 1

    10/18

    HOW ANN WORK?

    In the preceding figure, all of the zero th inputs to either thehidden our output la yer are referred to as thresholds and arety picall y set to -1.

    The weights of a neural network can be an y positive or negative value.

    The input values are multiplied b y the weights that connect

    them to a particular neuron.Neurons take this weighted sum as input and use an

    activation function to compute the neurons output.

    The output of one neuron becomes the input to another neuron multiplied b y a different subset of weights.

  • 8/8/2019 Artificial Neural Networking 1

    11/18

    TYPE S OF NETWORK

    Multilayer Perceptron

    Radial Basis Function

    Kohonen

    Linear

    Hopfield

    Adaline/Madaline

    Probabilistic Neural Network (PNN)General Regression Neural Network (GRNN)

    and at least thirty others

  • 8/8/2019 Artificial Neural Networking 1

    12/18

    NEURAL NETWORK S US ES

    S peech recognition S peech synthesis Im age recognition

    Pattern recognition S tock m arket predictionRobot control and navigation

  • 8/8/2019 Artificial Neural Networking 1

    13/18

    Strengths of Artificial Neural NetworksNeural NetworksAre Versatile

    Neural Networks Are Versatile Neural Networks Can Produce Good

    Results in Complicated Domains

    Neural Networks Can HandleCategorical and Continuous Data T y pes

    Neural Networks Are Available inMan y Off-the- S helf Packages

    S TRENGTH S OF NEURAL NETWORKING

  • 8/8/2019 Artificial Neural Networking 1

    14/18

    All Inputs and Outputs Must BeMassaged to

    Neural Networks Cannot Explain

    Results Neural Networks Ma y Converge on

    an Inferior S olution

    WEAKNE SS ES OF ARTIFICIAL

    NEURAL NETWORK S

  • 8/8/2019 Artificial Neural Networking 1

    15/18

    CONCLUSION Neural network are ver y flexible and powerful.

    If used sensibl y they can produce some amazingresults.

    It has a ver y vast scope in this modern world.

  • 8/8/2019 Artificial Neural Networking 1

    16/18

    REFERENCE S

    B . YEGNANARAYANA, S ur yakanth V. Gangashett y, andS . Palanivel, Autoassociative Neural Network Models for Pattern Recognition Tasks in S peech and Image, in AshishGhosh and S ankar K. Pal (Eds.), S oft ComputingApproach to Pattern Recognition and Image Processing,World S cientific Publishing Co. Pte. Ltd., S ingapore,2002.

    B . YEGNANARAYANA and C. Chandra S ekhar, PatternRecognition Issues in S peech Processing in S ankar. K. Paland Amita Pal (Eds.), Pattern Recognition from Classicalto Modern Approaches, World S cientific, S ingapore, 2001.

  • 8/8/2019 Artificial Neural Networking 1

    17/18

  • 8/8/2019 Artificial Neural Networking 1

    18/18