Indian Institute of Technology Kanpur
IME Department
IME673: Applied Machine Learning
The course will help you learn to build predictive models and own them. You
can own the model if you have the domain knowledge, understand the data,
have explored the model for generalizability and have confidence in the models
that you build. Please note that I will not share any PPTs. You are expected to
take notes in the class. There will be a demonstration of each concept using
python programs. We will share these programs. Attendence carries no weight.
The course will have the following components.
(1) Predictive model building and related assignments using Python
(2) Mid Sem exam
(3) End Sem exam
Course Contents
S. No. Broad Title Topics
1. Introduction A. Why Ml, why now
B. Well posed ML problem
C. Examples of Machine Learning with data
requirements and technique used
D. Data
E. Forms of Learning
F. Learning from Observations
G. Bias and Variance
H. Why learning works
I. Occam’s razor principle
J. Heuristic Search in inductive learning
2. Exploratory Data A. Univariate data analysis: Variance, Skewness,
Analysis distribution, Missing Values, Outliers
B. Multivariate data analysis: Correlations,
Relationships among predictors and target
variable
3. Linear Regression A. Linear Regression: with minimum sum of
error square and pseudoinverse
B. Gradient descent optimization
C. Least mean square algorithm
4. Performance Measure C. Generalization, overfitting, underfitting
D. Class imbalance problem
E. Accuracy, Precision, Recall, AUROC, f1-score etc.
5. Classification A. Naïve Bayes Classifier with example
B. K-nearest neighbour
C. Logistic regression for classification
6. Case Discussion To demonstrate how ML is used in practice
7. Classification 1. Support Vector Machine
2. Linear Discriminant Function for Binary
Classification
3. Linear Maximal Margin Classifier for Linearly
Separable classes
4. Linear Soft Margin Classifier for Overlapping
classes
5. Multiclass classification
8. Classification A. Towards cognitive machine
B. Perceptron Algorithm
C. Neuron Models
D. Network Architecture
E. Linear classification using regression
techniques
F. Standard Gradient Descent Optimization
scheme: we have already discussed it, tie it
together
G. Linear Neuron & Widrow – Hoff learning rule
H. Error Correction Delta rule
I. Multi-layer perceptron (MLP) network and the
error propagation algorithm
J. Generalized delta rule: pick one activation rule
K. Convergence and local minima
L. Adding momentum to gradient descent
M. Heuristics
N. Multiclass Discrimination with MLP
O. Radial Basis Functions (RBF) network
9. Unsupervised Learning A. Introduction + Clustering
B. Clustering analysis
C. Overview of basic clustering algorithms
D. Hierarchical clustering
E. K-means clustering
10. Decision Tree Learning A. Decision Tree Learning
B. Example
C. Measure of impurity for evaluating splits
D. Gain Ratio
E. Gini Index
F. ID3, C4.5, CART
G. Pruning the tree
H. Strength & Weaknesses of decision-tree approach
12. Data Transformation A. Data Cleansing, Derived attributes etc.
1. Recommended books:
Text Book: M Gopal, Applied Machine Learning, McGraw Hill, 2018
Reference Books:
1. Max Kuhn and Kjell Johnson, Applied Predictive Modeling, Springer, 2013
2. John C. Hull, Machine Learning in Business: An Introduction to the World of
Data, Independently Published, 2021
2. Grading
A) Assignments: 15%
B) Quizes: 5%
C) Mid Sem Exam: 40%
D) End Sem Exam: 40%
