NP Completeness: Polynomial time reductions, Efficient Certification and

Definition of NP, NP Complete problems, Sequencing problems, Partitioning

12-13
problems, co-NP and asymmetry of NP

Backtracking: Constructing All Subsets, Constructing All Permutations,

14-15 Constructing All Paths in a Graph

Assessment Methods
Written tests, assignments, quizzes, presentations as announced by the instructor in the class.

Keywords
Algorithms, Analysis, Network Flows, NP Completeness.

Machine Learning (BHCS18A) Discipline Specific Elective - (DSE)

Credit: 06

Course Objective
The course aims at introducing the basic concepts and techniques of machine learning so that a
student can apply machine learning techniques to a problem at hand.

Course Learning Outcomes

On successful completion of this course, the student will be able to:
1. Differentiate between supervised and unsupervised learning tasks.
2. Differentiate between linear and non-linear classifiers.
3. Describe theoretical basis of SVM
4. Implement various machine learning algorithms learnt in the course.

Detailed Syllabus

91
Unit 1
Introduction: Basic definitions, Hypothesis space and inductive bias, Bayes optimal classifier
and Bayes error, Occam's razor, Curse of dimensionality, dimensionality reduction, feature
scaling, feature selection methods.
Unit 2
Regression: Linear regression with one variable, linear regression with multiple variables,
gradient descent, logistic regression, over-fitting, regularization. performance evaluation
metrics, validation methods.
Unit 3
Classification: Decision trees, Naive Bayes classifier, k-nearest neighbor classifier, perceptron,
multilayer perceptron, neural networks, back-propagation algorithm, Support Vector Machine
(SVM), Kernel functions.
Unit 4
Clustering: Approaches for clustering, distance metrics, K-means clustering, expectation
maximization, hierarchical clustering, performance evaluation metrics, validation methods.

Practical
For practical Labs for Machine Learning, students may use softwares like MABLAB/Octave or
Python. For later exercises, students can create/use their own datasets or utilize datasets from
online
repositories like UCI Machine Learning Repository (http://archive.ics.uci.edu/ml/).
1. Perform elementary mathematical operations in Octave/MATLAB like addition,
multiplication, division and exponentiation.
2. Perform elementary logical operations in Octave/MATLAB (like OR, AND, Checking for
Equality, NOT, XOR).
3. Create, initialize and display simple variables and simple strings and use simple formatting for
variable.
4. Create/Define single dimension / multi-dimension arrays, and arrays with specific values like
array of all ones, all zeros, array with random values within a range, or a diagonal matrix.
5. Use command to compute the size of a matrix, size/length of a particular row/column, load
data from a text file, store matrix data to a text file, finding out variables and their features in the
current scope.

92
6. Perform basic operations on matrices (like addition, subtraction, multiplication) and display
specific rows or columns of the matrix.
7. Perform other matrix operations like converting matrix data to absolute values, taking the
negative of matrix values, additing/removing rows/columns from a matrix, finding the maximum
or minimum values in a matrix or in a row/column, and finding the sum of some/all
elements in a matrix.
8. Create various type of plots/charts like histograms, plot based on sine/cosine function based on
data from a matrix. Further label different axes in a plot and data in a plot.
9. Generate different subplots from a given plot and color plot data.
10. Use conditional statements and different type of loops based on simple example/s.
11. Perform vectorized implementation of simple matrix operation like finding the transpose of a
matrix, adding, subtracting or multiplying two matrices.
12. Implement Linear Regression problem. For example, based on a dataset comprising of
existing set of prices and area/size of the houses, predict the estimated price of a given house.
13. Based on multiple features/variables perform Linear Regression. For example, based on a
number of additional features like number of bedrooms, servant room, number of balconies,
number of houses of years a house has been built – predict the price of a house.
14. Implement a classification/ logistic regression problem. For example based on different
features of students data, classify, whether a student is suitable for a particular activity. Based on
the available dataset, a student can also implement another classification problem like checking
whether an email is spam or not.
15. Use some function for regularization of dataset based on problem 14.
16. Use some function for neural networks, like Stochastic Gradient Descent or backpropagation
- algorithm to predict the value of a variable based on the dataset of problem 14.

References
1. Flach, P. (2015). Machine Learning: The Art and Science of Algorithms that Make Sense of
Data. Cambridge University Press.
2. Mitchell, T.M. (2017). Machine Learning. McGraw Hill Education.

Additional References:
1. Christopher & Bishop, M. (2016). Pattern Recognition and Machine Learning. New York:
Springer-Verlag

93
2. Haykins, S.O. (2010). Neural Networks and Learning Machines. 3rd edition. PHI.

Course Teaching Learning Process

● Use of ICT tools in conjunction with traditional class room teaching methods
● Interactive sessions
● Class discussions

Tentative weekly teaching plan is as follows:

Week Content
1 Basic definitions, Hypothesis space and inductive bias, Bayes optimal classifier
and Bayes error, Occam's razor
2 Curse of dimensionality, dimensionality reduction, feature scaling, feature
selection methods
3 Linear regression with one variable, linear regression with multiple variables

4 -5 Gradient descent, logistic regression, over-fitting, regularization

6 Performance evaluation metrics, validation methods.

7 Decision trees, Naive Bayes classifier ,k-nearest neighbor classifier

8 - 9 Perceptron, Multilayer perceptron, neural networks, back-propagation algorithm
10-11 Support Vector Machine (SVM), Kernel functions
12 Approaches for clustering, distance metrics
13 K-means clustering, expectation maximization
14 Hierarchical clustering
15 Clustering validation methods, performance evaluation metrics

Assessment Methods
Written tests, assignments, quizzes, presentations as announced by the instructor in the class.

Keywords
Machine learning, unsupervised learning, supervised learning, support vector machines, neural
networks, classification, clustering,

94