ann intro jan2010
TRANSCRIPT
-
8/11/2019 ANN Intro Jan2010
1/29
1
Artificial Neural Networks:An Introduction
S. Bapi Raju
Dept. of Computer and
Information Sciences,University of Hyderabad
-
8/11/2019 ANN Intro Jan2010
2/29
ANN-Intro (Jan 2010)
2 of 29
OUTLINEBiological Neural Networks
Applications of Artificial Neural Networks
Taxonomy of Artificial Neural Networks
Supervised and Unsupervised Artificial NeuralNetworks
Basis function and Activation function
Learning Rules
Applications OCR, Load Forecasting, Condition Monitoring
-
8/11/2019 ANN Intro Jan2010
3/29
ANN-Intro (Jan 2010)
3 of 29
Biological Neural NetworksStudy of Neural Networks originates in biological systems
Human Brain:contains over 100 billion neurons, number ofsynapses is approximately 1000 times that
in electronic circuit terms: synaptic fan-in fan-out is 1000,
switching time of a neuron is order of milliseconds
But on a face recognition problem brain beats fastest
supercomputer in terms of number of cycles of computationto arrive at answer
Neuronal Structure
Cellbody
Dendrites for inputAxon carries output to otherdendritesSynapse-where they meetActivation signal (voltage)travels along axon
-
8/11/2019 ANN Intro Jan2010
4/29
ANN-Intro (Jan 2010)
4 of 29
Need for ANNStandard Von Neumman Computing as existing
presently has some shortcomings.
Following are some desirable characteristics in ANN
Learning AbilityGeneralization and Adaptation
Distributed and Parallel representation
Fault Tolerance
Low Power requirementsPerformance comes not just from the computational
elements themselves but the manner of networkedinterconnectedness of the decision process.
-
8/11/2019 ANN Intro Jan2010
5/29
ANN-Intro (Jan 2010)
5 of 29
VonNeumann
versus
BiologicalComputer
-
8/11/2019 ANN Intro Jan2010
6/29
ANN-Intro (Jan 2010)
6 of 29
ANN Applications
Pattern Classification Speech Recognition, ECG/EEG classification, OCR
-
8/11/2019 ANN Intro Jan2010
7/29
ANN-Intro (Jan 2010)
7 of 29
ANN Applications
Clustering/Categorization Data mining, data compression
-
8/11/2019 ANN Intro Jan2010
8/29
ANN-Intro (Jan 2010)
8 of 29
ANN Applications
Function Approximation Noisy arbitrary function needs to be approximated
-
8/11/2019 ANN Intro Jan2010
9/29
ANN-Intro (Jan 2010)
9 of 29
ANN Applications
Prediction/Forecasting Given a function of time, predict the function values
for future time values, used in weather predictionand stock market predictions
-
8/11/2019 ANN Intro Jan2010
10/29
ANN-Intro (Jan 2010)
10 of 29
ANN Applications
Optimization Several scientific and other problems can be reduced
to an optimization problem like the TravelingSalesman Problem (TSP)
-
8/11/2019 ANN Intro Jan2010
11/29
ANN-Intro (Jan 2010)
11 of 29
ANN Applications
Content Based Retrieval Given the partial description of an object retrieve the
objects that match this
-
8/11/2019 ANN Intro Jan2010
12/29
f
-
8/11/2019 ANN Intro Jan2010
13/29
ANN-Intro (Jan 2010)
13 of 29
Characteristics of ANNBiologically inspired computational units
Also called as Connectionist Models orConnectionist Architectures
Large number of simple processing elementsVery large number of weighted connectionsbetween elements. Information in thenetwork is encoded in the weights learned by
the connectionsParallel and distributed control
Connection weights are learned by automatictraining techniques
14 f 29
-
8/11/2019 ANN Intro Jan2010
14/29
ANN-Intro (Jan 2010)
14 of 29
Artifical Neuron Working Model
Objective is to create a model of functioning ofbiological neuron to aid computation
All signals at synapses are
summed i.e. all the excitatory
and inhibitory influences and
represented by a net value h(.)
If the excitatory influences are
dominant, then the neuron fires,
this is modeled by a simple
threshold function (.)
Certain inputs are fixed biases
Output yleads to other
neurons
McCulloch Pitts Model
15 f 29
-
8/11/2019 ANN Intro Jan2010
15/29
ANN-Intro (Jan 2010)
15 of 29
More about the Model
Activation Functions play a key role Simple thresholding (hard limiting)
Squashing Function (sigmoid)
Gaussian Function Linear Function
Biases are also learnt
16 f 29
-
8/11/2019 ANN Intro Jan2010
16/29
ANN-Intro (Jan 2010)
16 of 29
Different Kinds of NetworkArchitectures
17 f 29
-
8/11/2019 ANN Intro Jan2010
17/29
ANN-Intro (Jan 2010)
17 of 29
Learning AbilityMere Architecture is insufficient
Learning Techniques also need to be formulated
Learning is a process where connection weights are
adjustedLearning is done by training from labeled examples.This is the most powerful and useful aspect of neuralnetworks in their use as Black Box classifiers.
Most commonly an input-output relationshipis learntLearning Paradigm needs to be specified
Weight update in learning rules must be specified
Learning Algorithm specifies step by step procedure
18 of 29
-
8/11/2019 ANN Intro Jan2010
18/29
ANN-Intro (Jan 2010)
18 of 29
Learning TheoryMajor Factors Learning Capacity: This concerns the number of
patterns that can be learntand the functionsand kinds of decision boundaries that can be
formed Sample Complexity: This concerns the number of
the samples needed to learn withgeneralization. Overfitting problem is to beavoided
Computational Complexity: This concerns thecomputation time neededto learn theconcepts embedded in the training samples.Generally the computational complexity of learning
is high.
-
8/11/2019 ANN Intro Jan2010
19/29
20 of 29
-
8/11/2019 ANN Intro Jan2010
20/29
ANN-Intro (Jan 2010)
20 of 29
Major Learning Rules
Error Correction:Error signal (dy)used to adjust the
weights so thateventually desiredoutput disproduced
Perceptron SolvingAND Problem
21 of 29
-
8/11/2019 ANN Intro Jan2010
21/29
ANN-Intro (Jan 2010)
21 of 29
Major Learning Rules
Error Correction: in Mutlilayer Feedforward Network
Geometric interpretation of the role of hidden units in a 2D input space
22 of 29
-
8/11/2019 ANN Intro Jan2010
22/29
ANN-Intro (Jan 2010)
22 of 29
Major Learning RulesHebbian:weights are adjusted by afactor proportional to the activities ofthe neurons associated
OrientationSelectivity of aSingle HebbianNeuron
23 of 29
-
8/11/2019 ANN Intro Jan2010
23/29
ANN-Intro (Jan 2010)
23 of 29
Major Learning Rules
Competitive Learning: winner take all
(a) Before Learning (b) After Learning
24 of 29
-
8/11/2019 ANN Intro Jan2010
24/29
ANN-Intro (Jan 2010)
24 of 29
Summary of ANN Algorithms
25 of 29
-
8/11/2019 ANN Intro Jan2010
25/29
ANN-Intro (Jan 2010)
25 of 29
26 of 29
-
8/11/2019 ANN Intro Jan2010
26/29
ANN-Intro (Jan 2010)
26 of 29
Application to OCR SystemThe main problemin the HandwrittenLetter recognition isthat characters withvariation inthickness shape,rotation anddifferent nature ofstrokes need to berecognized as ofbeing in thedifferent categoriesfor each letter.
Sufficient number ofsample training datais required for eachcharacter to train
the networks
A Sample set of characters in the NIST Data
27 of 29
-
8/11/2019 ANN Intro Jan2010
27/29
ANN-Intro (Jan 2010)
27 of 29
OCRProcess
28 of 29
-
8/11/2019 ANN Intro Jan2010
28/29
ANN-Intro (Jan 2010)
28 of 29
OCR Example (continued)
Two schemes shown at right
First makes use of the
feature extractors
Second uses the image
pixels directly
29 of 29
-
8/11/2019 ANN Intro Jan2010
29/29
ANN-Intro (Jan 2010)
References
A. K. Jain, J.Mao, K.Mohiuddin, ANN aTutorial, IEEE Computer, 1996 March, pp 31-44 (Figures and Tables taken from this reference)
B. Yegnanarayana,Artificial Neural Networks,Prentice Hall of India, 2001.
Y. M. Zurada, Inroduction toArtificial Neural
Systems, Jaico, 1999.MATLAB neural networks toolbox and manual