Mehdi Ghayoumi MSB rm 132 [email protected] Ofc hr: Thur, 11-12 a Machine Learning.

Mehdi Ghayoumi

MSB rm 132

[email protected]

Ofc hr: Thur, 11-12 a

Machine Learning

Machine Learning

THE NAÏVE BAYES CLASSIFIER

In the naïve Bayes classification scheme, the required

estimate of the pdf at a point x=[x(1),...,x(l)]T R∈ l is given as

That is, the components of the feature vector x are assumed

to be statistically independent.

Machine Learning

Example .

Generate a set X1 that consists of N1 = 50 5-dimensional data

vectors that stem from two classes, ω1 and ω2. The classes are

modeled by Gaussian distributions with means

m1 = [0,0,0,0,0]T and m2 = [1,1,1,1,1]T and respective covariance matrices

Machine Learning

Step 1. Classify the points of the test set X2 using

the naive Bayes classifier, where for a given x, p(x|

ωi ) is estimated as

Machine Learning

Step 2. Compute the ML estimates of m1, m2, S1,

and S2 using X1. Employ the ML estimates in the

Bayesian classifier

Machine Learning

Step 3. Compare the results obtained in steps 1 and 2.The resulting classification errors—naive_error and

Bayes_ML_error—are 0.1320 and 0.1426,

respectively.

In other words, the naive classification scheme

outperforms the standard ML-based scheme.

Machine Learning

The techniques that are built around the optimal

Bayesian classifier rely on the estimation of the pdf

functions describing the data distribution in each

class.

Machine Learning

Machine Learning

Machine Learning

Machine Learning

The focus is on the direct design of a discriminant

function/decision surface that separates the classes

in some optimal sense according to an adopted

criterion.

Machine Learning

Machine learning involves adaptive mechanisms that enable computers to Machine learning involves adaptive mechanisms that enable computers to

learn from experience, learn by example and learn by analogy. Learning learn from experience, learn by example and learn by analogy. Learning

capabilities can improve the performance of an intelligent system over capabilities can improve the performance of an intelligent system over

time. The most popular approaches to machine learning are time. The most popular approaches to machine learning are artificial artificial

neural networks neural networks and and genetic algorithmsgenetic algorithms. This lecture is dedicated to . This lecture is dedicated to

neural networks.neural networks.

• Cell structures– Cell body– Dendrites– Axon– Synaptic terminals

Machine Learning

• Networks of processing units (neurons) with connections (synapses)

between them

• Large number of neurons: 1010

• Large connectitivity: 105

• Parallel processing

• Distributed computation/memory

• Robust to noise, failures

Machine Learning

Understanding the Brain• Levels of analysis (Marr, 1982)

1. Computational theory

2. Representation and algorithm

3. Hardware implementation

• Reverse engineering: From hardware to theory

• Parallel processing: SIMD vs MIMD

Machine Learning

Real Neural Learning

• Synapses change size and strength with experience.

• 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.”

Machine Learning

Machine Learning

Soma Soma

Synapse

Synapse

Dendrites

Axon

Synapse

Dendrites

Axon

Machine Learning

BiologicalNeuralNetwork Artificial NeuralNetworkSomaDendriteAxonSynapse

NeuronInputOutputWeight

Neural Network Learning

• Learning approach based on modeling adaptation in

biological neural systems.

• Perceptron: Initial algorithm for learning simple neural

networks (single layer) developed in the 1950’s.

• Backpropagation: More complex algorithm for learning

multi-layer neural networks developed in the 1980’s.

Machine Learning

Perceptron Learning Algorithm

• First neural network learning model in the 1960’s

• Simple and limited (single layer models)

• Basic concepts are similar for multi-layer models so this

is a good learning tool

• Still used in many current applications

Machine Learning

Machine Learning

Neuron Y

InputSignals

x1

x2

xn

OutputSignals

Y

Y

Y

w2

w1

wn

Weights

Machine Learning

Step function Sign function

+1

-1

0

+1

-1

0X

Y

X

Y

1 1

-1

0 X

Y

Sigmoid function

-1

0 X

Y

Linear function

0if,0

0if,1

X

XYstep

0if,1

0if,1

X

XYsign X

sigmoid

eY

1

1XYlinear

Machine Learning

Threshold

Inputs

x1

x2

Output

Y

HardLimiter

w2

w1

LinearCombiner

Perceptron Node – Threshold Logic Unit

x1

xn

x2

w1

w2

wn

z

Machine Learning

x1

xn

x2

w1

w2

wn

z

• Learn weights such that an objective

function is maximized.

• What objective function should we use?

• What learning algorithm should we use?

Machine Learning

Perceptron Learning Algorithm

x1

x2

z

.4

-.2

.1

Machine Learning

First Training Instance

.8

.3

z

.4

-.2

.1

net = .8*.4 + .3*-.2 = .26

=1

Machine Learning

Second Training Instance

.4

.1

z

.4

-.2

.1

net = .4*.4 + .1*-.2 = .14

=1

Machine Learning

Thank you!