Machine Learning Techniques · IIT Madras BS Degree

Lecture notes for all 74 lectures

Notes for Prof. Arun Rajkumar's course, grouped into 18 parts by topic. Each part has diagrams of the key ideas, the maths written out, timestamp links into the videos, and questions to test yourself.

Start with Part 1 Open the playlist

Course map

Parts follow topic boundaries, so some cover two lectures and some cover nine.

PartTopicLecturesNotes
Introduction
01Introduction and paradigmsL1–L2Ready
Unsupervised learning
02Representation learningL3–L5Ready
03Principal component analysisL6–L7Coming
04Issues with PCA, and computing it fastL8–L9Coming
05Feature maps, kernels and kernel PCAL10–L12Coming
06K-means: the algorithm and its convergenceL13–L15Coming
07K-means: geometry, initialization, choosing KL16–L18Coming
08Estimation: maximum likelihood and BayesianL19–L21Coming
09Gaussian mixture models and EML22–L26Coming
Supervised learning
10Linear regression and gradient descentL27–L31Coming
11Kernel and probabilistic regressionL32–L35Coming
12Ridge and LASSOL36–L40Coming
13Classification: k-NN and decision treesL41–L44Coming
14Generative models and Naive BayesL45–L51Coming
15Perceptron and logistic regressionL52–L56Coming
16Support vector machinesL57–L65Coming
17Ensembles: bagging and boostingL66–L68Coming
18Loss functions, neural networks and summaryL69–L74Coming

How the course is laid out

The course starts with unsupervised learning and then moves to supervised learning. Sequential learning (online learning, bandits, reinforcement learning) is mentioned in Part 1 but not taught.

Unsupervised Supervised Start (L1–L2) Finish RepresentationlearningL3–L12 ClusteringL13–L18 EstimationL19–L26 BasicalgorithmsL27–L44 AdvancedalgorithmsL45–L65 AdvancedtopicsL66–L74
The course in six levels. The level names are the lecturer's. The lecture ranges are mapped from the playlist titles for these notes.

Using these notes

  • Every section starts with a timestamp chip, such as L6 · 12:40, that opens the video at that moment.
  • Each part ends with key terms and questions whose answers stay hidden until you open them.
  • Charts built on synthetic data say so in their caption. They illustrate the idea, not a dataset from the course.
  • The notes follow the lectures, not a textbook. For more depth the lecturer recommends Bishop's Pattern Recognition and Machine Learning.