ES 654 IIT Gandhinagar

Home Lectures Assessment Recordings

Lectures

Every class of the 2020 offering, numbered in order, with its slides and recording. Lectures ran from January to July 2020.

SlidesFull recordingReading

Foundations and decision trees1–5Linear regression and optimisation6–13Bayesian and probabilistic models14–20Neural networks and support vector machines21–28Naive Bayes, nearest neighbours and further topics29–32

Full lecture playlist

Foundations and decision trees

Lectures 1–5

Lecture 01	Introduction	Slides Recording
Lecture 02	Conventions and metrics for classification and regression	Slides Recording
Lecture 03	Decision trees I	Slides Recording
Lecture 04	Decision trees II; bias and variance	Slides Recording
Lecture 05	Ensemble methods	Slides Recording

Linear regression and optimisation

Lectures 6–13

Lecture 06	Linear regression I; maths for ML I	Slides: Linear regressionSlides: Maths for ML I Recording
Lecture 07	Geometric interpretation of least squares; linear regression II; contour plots	Slides: Geometric interpretationSlides: Linear regression IISlides: Contour plots Recording: Geometric interpretationRecording: Linear regression IIRecording: Contour plots
Lecture 08	Gradient descent	Slides Recording
Lecture 09	Convexity; causality and fairness Causality and fairness: guest lecture by Ritik Dutta.	Slides: ConvexitySlides: Causal inference (Ritik Dutta) Recording: ConvexityRecording: Causality and fairness
Lecture 10	Time complexity of linear regression; constrained optimisation; decision trees with weighted samples	Slides: Time complexitySlides: Constrained optimisationSlides: Weighted decision trees Recording: Time complexityRecording: Constrained optimisationRecording: Weighted decision trees
Lecture 11	Constrained optimisation II; ridge regression	Slides: Ridge regression Recording: Constrained optimisation IIRecording: Ridge regression
Lecture 12	Ridge regression II; lasso regression	Slides: Lasso regression Recording: Ridge regression IIRecording: Lasso regression
Lecture 13	Lasso regression II; multivariate normal distribution	Slides: Multivariate normal Recording: Lasso regression IIRecording: Multivariate normal

Bayesian and probabilistic models

Lectures 14–20

Lecture 14	Bayesian learning I; Bayesian linear regression	Slides: Bayesian learningSlides: Bayesian linear regression Recording: Bayesian learningRecording: Bayesian linear regression
Lecture 15	Bayesian linear regression II	Recording
Lecture 16	Bayesian linear regression III; logistic regression I and II	Slides: Logistic regression Recording: Bayesian linear regression IIIRecording: Logistic regression IRecording: Logistic regression II
Lecture 17	Hidden Markov models	Slides Recording, part 1Recording, part 2Recording, part 3Recording, part 4
Lecture 18	Entropy and differential entropy	Slides: EntropySlides: Differential entropy Recording, part 1Recording, part 2
Lecture 19	Gaussian processes I and II	Slides Recording: Gaussian processes IRecording: Gaussian processes II
Lecture 20	Gaussian processes: learning parameters	Recording

Neural networks and support vector machines

Lectures 21–28

Lecture 21	Multilayer perceptrons I	Slides Recording
Lecture 22	Multilayer perceptrons II	Recording
Lecture 23	Neural networks: backpropagation	Recording
Lecture 24	Convolutional neural networks	Slides Recording
Lecture 25	Support vector machines: introduction	Slides Recording
Lecture 26	Support vector machines: kernels	SlidesReading: Jake VanderPlas, Python Data Science Handbook, SVMs Recording
Lecture 27	Support vector machines: soft margin; Lagrangian and duality	Slides: Soft marginSlides: Lagrangian and duality Recording: Soft marginRecording: Lagrangian and duality
Lecture 28	Automatic differentiation	Slides Recording

Naive Bayes, nearest neighbours and further topics

Lectures 29–32

Lecture 29	Naive Bayes	Slides Recording
Lecture 30	K-nearest neighbours	Slides Recording
Lecture 31	Unsupervised learning	Slides Recording
Lecture 32	Reinforcement learning Held in February; numbered last as on the original page.	Recording

Slides are the PDFs used in class; recordings are on YouTube. Where a class covered several topics, each part has its own recording.