GAI10SG194 Pattern Recognition and Anomaly Detection L T P C
Version 1.0 3 0 0 3
Pre-requisites/Exposure Data Mining & Machine Learning
Co-requisites --
Course Objectives
1. To understand fundamentals of pattern and its computational significance.
2. To study various approaches through which pattern recognition can be carried out
Course Outcomes
On completion of this course, the students will be able to
CO1. Explain the basic concepts of pattern and anomaly from information theory
CO2. Demonstrate the steps involved in extraction and detection of data patterns
CO3. Practice the various models used for pattern recognition using to solve real world problems
approaches
Catalog Description
The course is aimed at specific problem, pattern recognition that is often encountered in data processing
applications. The course details of pattern and anomaly as a distinguishable feature and specifications
of them from information theory view point. Various methods that are available under statistical and
learning approaches through which pattern analysis especially from recognition and detection purpose
is addressed in this course.
Course Content
Unit 1. Pattern and Anomaly Detection Introduction 09 Lecture Hours
What is pattern? What is pattern recognition? Pattern recognition techniques, Training and learning
in pattern recognition, Pattern recognition applications, Pattern recognition use cases, what is
anomaly detection? What are some other practical uses for anomaly detection? How is anomaly
detection calculated over time? Key point for AI and ML-anomaly detection, Tasks for artificial
intelligence, AI system learning process, Test to geometric requirements for curves algebraic, Curves
matched to data points, Case study: Anomaly detection with IBM Watson, Probability theory,
Maximum likelihood theory and estimation, Model selection, Matrices of uncertainty (confusion
matrices), Loss of logging (log-loss), Rate for F1 (F1 score), Metric selection, Hyperparameter selection,
The problem with high dimensionality, Information theory.
Unit 2. Statistical Approaches for Pattern Recognition 09 Lecture Hours
Understanding statistics, T-test, Z-test, Z-test and t-test difference, P-value, Descriptive statistics, Type
I error, Type II error, Differences between type I and type II errors, Null hypothesis, Statistical
significance, Hypothesis testing, Four steps of hypothesis testing, Real-world example of hypothesis
testing, Bonferroni test, Check of one-tailed, Probability distributions, Types of distributions,
Regression models, Types of regression, How to select the best model for regression? Common
questions, Linear models for classification, Example of positive linear regression.
Unit 3. Machine Learning Approaches for Pattern Recognition 09 Lecture Hours
Neural networks, how neural networks learn? Neural networks examples, Neural networks use cases,
Kernel methods, Sparse kernel machines use cases, Graphical models, Mixture models and EM,
Bayesian networks: Directed graphical models, Conditional probability distributions, Potential
functions, Conditional independences, Sampling methods for pattern recognition, Continuous latent
variables, Combining models for pattern recognition, Markov chain Monte Carlo, The K-means
algorithm, Applications of K-means.
Unit 4. Anomaly Detection & Anomaly Detection Approaches 09 Lecture Hours
What are anomalies? Applications of anomaly detection, Related use cases, Types of input data, Types
of anomalies, Evaluation of an anomaly detector, Taxonomy of approaches, Classification based,
Classification use cases, Supervised classification techniques, Nearest neighbour based techniques,
Others model techniques, Information theory, Contextual anomaly based, Collective anomaly
detection, On-line based model, Distributed anomaly detection, IDS analysis strategy.
Unit 5. Real-world problems 09 Lecture Hours
Network intrusion detection, Understanding of IDS core operation, How an IDS works? Types of
intrusion detection systems, Fundamental concerns of intrusion detection systems, Intrusion
detection vs. intrusion prevention, The future of IDS, Anomaly detection in big data, Key attributes of
advanced anomaly detection, The real-world impact of anomaly detection, Anomaly detection on 5G:
Possibilities and opportunities, Real time anomaly detection in docker, Hadoop cluster, Anomaly
detection in IoT, Detection of deviations in deep learning time series results, Anomaly detection use
cases, Anomaly detection with time series forecasting, What is time series analysis? Time series data
models, how to find anomaly in time series data? Anomaly detection using machine learning, Anomaly
detection using deep learning, Anomaly detection for an e-commerce pricing system, IBM’s Watson
AIOps automates IT anomaly detection and remediation.
Text Book - Pattern and Anomaly Detection (IBM ICE Publications)
Reference Materials
1. Richard O. Duda, Peter E. Hard, David G. Stork, Pattern Recognition, 2nd, Wiley,
2021.
Modes of Evaluation: Quiz/Assignment/ presentation/ extempore/ Written Examination
Examination Scheme:
Components MSE Presentation/Assignment/ etc ESE
Weightage (%) 20 % 30 % 50 %
Course Outcomes PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10 PO11 PO12 PSO1 PSO2 PSO3
CO1 1 3
CO2 1 2 3
CO3 2 3
Average 1 2 3
1=weak 2= moderate 3=strong
