Lecture 22: CS217 | SVM Implementation & Introduction to Hidden Markov Models | AI-ML | IITB 2025

Welcome to Lecture 22 of the CS217: AI-ML Course by IIT Bombay. In this session, Prof. Pushpak Bhattacharya begins with a review of previous concepts before guiding students through the practical implementation of Support Vector Machines using scikit-learn, and introducing Hidden Markov Models (HMMs) as a powerful technique for sequence labeling tasks.

Topics Covered:

SVM Implementation with Scikit-learn - Detailed overview of implementing SVMs using Python's sklearn package, including data handling with pandas, attribute normalization using standard scalar, and evaluating model performance through precision, recall, and F-score metrics. Discussion of how these evaluation metrics are mathematically defined and their practical implications in machine learning.
Precision, Recall & F-Score - In-depth explanation of evaluation metrics for binary classification, visualizing the relationship between true positives, false positives, false negatives, and true negatives. Special focus on the harmonic mean (F-score) and why it's preferred over arithmetic mean when combining precision and recall.
Introduction to Hidden Markov Models - Foundation concepts of HMMs for sequence labeling problems, including the noisy channel model and applications in computer vision (posture recognition) and natural language processing (part-of-speech tagging). Presentation of the classic urn-and-balls example to illustrate hidden states and observations.
HMM Algorithms Overview - Introduction to the three classic problems in HMM: finding the most likely state sequence (Viterbi algorithm), computing observation sequence probability (Forward-Backward algorithm), and parameter estimation (Baum-Welch algorithm).

This lecture bridges theoretical concepts with practical implementation while introducing a new probabilistic framework for handling sequential data. It offers essential guidance for students preparing for their upcoming lab assignment on SVMs while laying the groundwork for future sessions on Hidden Markov Models.

\#supportvectormachines #svm #sklearnimplementation #hiddenmarkovmodels #hmm #machinelearning #aiml #iitbombay #computerscience #sequencelabeling #precision #recall #fscore #evaluationmetrics #viterbi #baumwelch #noisychannel #probabilisticmodels #cs217 #aimlcourse #posrecognition #partofspeechtagging #classifiers #computerscience #machinelearningtheory #aiml #iitb #chatgpt #anthropic #deepseek #grok