Introduction to Pattern Recognition 1 Lecture 1 Dr. M. Aksam Iftikhar, Associate Professor (CS) Syllabus ▪ Prerequisite ▪ First - year course in Calculus ▪ Introductory Probability/ Statistics ▪ Some Linear Algebra ▪ Grading ▪ Midterm ▪ Assignments ▪ Research Project/Report will re v i e w and provide related materials 2 Syllabus ▪ Assignments ▪ T heoretical or programming in Python ▪ Applications of discussed algorithms. ▪ Submitted on Google Classroom or by - hand ▪ Midterm ▪ Conceptual questions reflecting understanding of the student 3 Syllabus ▪ Research Project/Report (more about this later) ▪ Choose from the list of topics or design your own 4 Course Details Google Classroom Code: szwaf5w OR joining link: https://classroom.google.com/u/1/c/MzkwODI0ODY2NzA5 ( Note: Join using CUI - Lahore email ID ) % 9 ©1992 - 2012 by Pearson Education, Inc. All Rights Reserved. Pre - Requisites Good Programming Skills 11 Lecture Outline ▪ About Me ▪ What is pattern and pattern recognition? ▪ Some terminologies & interesting applications ▪ Key ideas in pattern recognition with example ▪ Representation of patterns with feature vector ▪ Understanding class separability to identify good/bad features ▪ Understanding model complexity and related concepts ▪ Generalization, overfitting, underfitting, decision boundary, classification cost Lecture Outline (Contd.) ▪ Structure of a pattern recognition system ▪ Understanding with an example ▪ Design choices of a pattern recognition system ▪ Modes of pattern recognition with examples (supervised, unsupervised, reinforcement) ▪ Tentative Course Outline What is “ pattern ” “ recognition ” – some terminologies • A pattern is an object, process or event that can be given a name. • Object: Fingerprint/facial/iris Recognition, OCR, object detection (e.g. detecting a car/person in a video) • Process: Recognizing patterns in the progression of diseases can help in early diagnosis and treatment, Processes like sentence parsing involve recognizing structural patterns in language. • Event: Detecting unusual patterns of network events can signal a cyber - attack, Identifying patterns that precede earthquakes or volcanic eruptions. Some formal descriptions Pattern recognition: “The assignment of a physical object or event to one of several prespecified categories” -- Duda & Hart An informal definition: “A pattern is the opposite of a chaos ; it is an entity vaguely defined, that could be given a name.” (Watanabe) What is “ pattern ”? Face with emotions Finger prints 10 Examples on patterns from different domains... 4 More about Patterns • Following are examples of patterns from some other categories: Temporal Patterns 1. Stock Market Trends : Recognizing patterns in stock market data over time can be critical for trading strategies. 2. Weather Patterns : Meteorologists use pattern recognition to identify recurring climate conditions. 3. Speech Recognition : Understanding the pattern of sound waves over a time period to convert spoken language into written text. More about Patterns • Following are examples of patterns from some other categories: Behavioral Patterns 1. User Behavior in Websites : Identifying common navigation patterns can help improve user experience and increase sales for e - commerce websites. 2. Animal Migration Patterns : Studying the regular journeys undertaken by many species of animals. 3. Social Behavior : Recognizing patterns in social interactions can be useful in areas like sociology, psychology, and even in surveillance systems. More about Patterns • Following are examples of patterns from some other categories: Multi - Dimensional Patterns Patterns can also exist in multiple dimensions simultaneously. For example: 1. Video Recognition : This involves recognizing patterns in both the spatial arrangement of pixels and their temporal progression, such as detecting a fire incident from a live camera video 2. Multi - Sensor Data Fusion : Combining data from multiple sensors to recognize complex events or conditions, e.g., combining multiple medical evidence (MRI, clinical tests, patient age etc.) to diagnose a disease What is “ pattern ” “ recognition ” – some terminologies • A pattern class (or category) is a set of patterns sharing common attributes. Each pattern class is typically defined by a set of features that describe the patterns within that class. • Examples Example 1: Classifying Emails (Spam or Not Spam) • Pattern Classes : Spam, Not Spam • Features : Frequency of certain words, number of attachments, sender reputation, etc. Example 2: Medical Diagnosis (Healthy, Diseased) • Pattern Classes : Healthy, Diseased • Features : Blood pressure, cholesterol level, age, gender, etc. What is “ pattern ” “ recognition ” – some terminologies • Classification / Recognition • In the context of pattern recognition, classification refers to the process of categorizing or labeling an object into one of several predefined classes or groups based on its features. • This is why it’s called pattern ‘recognition’ • The goal is to build a model that accurately assigns new, unseen objects to one of these classes, ideally in a way that is both robust and interpretable. • A classifier is a n algorithm or (mathematical) model which performs the classification. • It takes an object (described by its feature vector) as input and outputs a class label. • Classifiers can be as simple as a set of heuristic rules or as complex as deep neural networks. 7 What is Pattern Recognition ? ▪ Informally ▪ Recognize patterns in data ▪ More formally ▪ Assign an object or an event to one of the several pre - specified categories (a category is usually called a class ) ▪ The goal of a PR system is to recognize patterns accurately face p h on e tea cup Application: male or female? Perfect PR system m al e f e m al e Output : c l ass e s 2 4 Input : Image Objects (pictures)