HIDDEN MARKOV
MODELS
Prof. Navneet Goyal
Department of Computer Science
BITS, Pilani
Presentation based on:
Topics
Markov Models
Hidden Markov Models
HMM Problems
Markov Analysis
A technique that deals with the
probabilities of future occurrences by
a

Unsupervised Learning
- Clustering
Prof. Navneet Goyal
BITS, Pilani
What is Cluster Analysis?
Finding groups of objects such that the objects in a group
will be similar (or related) to one another and different
from (or unrelated to) the objects in other

1 EXPLORING DATA
1.07 Z-scores
What you see here is the so-called tattoo density of football players, expressed in the percentage
of the body covered with tattoos. The dot plots and the standard deviations show that there is much
more variability in the d

1 EXPLORING DATA
1.03 Graphs and shapes of distributions
Researchers often want to summarize the data they have. They can do that, for instance, by means
of a frequency table. In this video I will show you how you can use frequency tables to build
informa

1 EXPLORING DATA
1.04 Mode, median and mean
Next to summarizing a distribution by means of graphs, it can also be useful to describe the center of
your distribution. There are three main ways in which you can do that. By means of the mode, the
median and

1 EXPLORING DATA
1.06 Variance and standard deviation
Tattoos are increasingly popular among football players. Imagine you want to know how much of
their bodies football players cover with tattoos. The dot plots you see here represent the
distributions of

1 EXPLORING DATA
1.01 Cases, variables and levels of measurement
Imagine youre very interested in football you know, that sport that some of us like to call
soccer. You are that person who wants to know all the details, like: how many goals were scored
by

1 EXPLORING DATA
1.02 Data matrix and frequency table
If youre conducting a study, it makes sense to think about your data in terms of cases and variables.
Cases are the persons, animals or things youre studying, and variables are the characteristics of
i

1 EXPLORING DATA
1.08 Example
Say I live in a city with 8 high schools. I want to know what, per high school, the average grade for
chemistry is. The lowest possible grade is a 0 and the highest possible grade is a 10. This is the data
matrix. You can see

Linear Classification Models:
Generative
Prof. Navneet Goyal
CS & IS
BITS, Pilani
Probabilistic Models
Probabilistic view of classification!
Models with linear decision boundaries arise
from simple assumptions about the
distribution of data
Two approaches

Logistic Regression
Prof. Navneet Goyal
CS & IS
BITS, Pilani
Logistic Regression
In Linear regression, the dependent variable is continuous
What if the dependent is dichotomous or binary?
A person will vote for Reagan (1) or Carter (0)?
A woman will give

STATISTICS AND LINEAR ALGEBRA PRELIMINARIES
Prof. Navneet Goyal
CS & IS Department
BITS, Pilani
Material in the presentation adapted from A Tutorial
on Principal Component Analysis by Lindsay I Smith
Statistics
Population & Sample
Relationship between ind

Linear Regression &
Classification
Prof. Navneet Goyal
CS & IS
BITS, Pilani
Topics
Modeling
Predictive Modeling
Regression
Classification
K-NN
SVM
Linear Models for Regression
Bias-variance Decomposition
Fundamentals of Modeling
Abstract representation of

BIRLA INSTITUTE OF TECHNOLOGY & SCIENCE, PILANI
BITS C464 MACHINE LEARNING
I Semester 2014-2015
WORKSHEET #5
MultiLayer Perceptron Model using Backpropogation
OBJECTIVE: Multilayer Layer Perceptron model
Learning a digit recognizer using Artificial Neura

R EIN FO R C EM EN T
LEA R N IN G
Navneet Goyal
Slides developed using material from:
Reinforcement Learning:An Introduction
Richard S. Sutton and Andrew G. Barto
MIT Press
&
Machine Learning by Stephen Marsland, CRC Press (Chapter
13)
Introduction
Reinf

Preliminaries
Prof. Navneet Goyal
CS & IS
BITS, Pilani
Topics
Probability Theory
Decision Theory
Information Theory
Topics
Probability Theory
Decision Theory
Information Theory
Probability Theory
Key concept is dealing with uncertainty
- Due to noise on m

Some new Developments in
Machine Learning
Navneet Goyal
Department of Computer Science, BITS-Pilani, Pilani
Campus, India
New IBM Chip - TrueNorth
Research published in Journal Science
Consists of electronic Neurons
It attempts to mimic the way brain reco

Perceptrons
Navneet Goyal, BITS-Pilani
Perceptrons
Labeled data is called Linearly Separable
Data (LSD) if there is a linear decision
boundary separating the classes
Perceptron achieves perfect separation on
LSD
Perceptron iterates over the training se

1 EXPLORING DATA
1.05 Range, interquartile range and box plot
As you might have noticed, tattoos are increasingly popular among football players. The so-called
tattoo sleeve in particular is rising on the football fields. A tattoo sleeve is what the name