ES 654 IIT Gandhinagar

Home Lectures Assessment Notebooks FAQ Slack

Lectures

Each lecture lists its slides and reading first, then videos, then practice: notebooks and interactive demos. Slides are the PDFs used in class; the notebooks are also collected on the Notebooks page.

SlidesNotebookInteractive labCodeReading

Foundations1–2Trees, model fit and ensembles3–7Linear models and optimisation8–17Neural networks18–23Naive Bayes, neighbours and clustering24–26Support vector machines27–29

Foundations

Lectures 1–2

Lecture 01	Introduction and logistics	Slides
·	Prerequisite test An open-book test on Python, linear algebra, probability and calculus, required to take the course.	Prerequisite test
Lecture 02	Conventions, metrics, classification and regression	Slides

Trees, model fit and ensembles

Lectures 3–7

Lecture 03	Decision trees I	Slides Notebook: Misc tips
Lecture 04	Decision trees II	Slides Notebook: Pandas tips
Lecture 05	Bias and variance	Slides Notebook: Python utilitiesNotebook: Grid search
Lecture 06	Bias and variance II, cross-validation	Slides
Lecture 07	Ensemble methods	Slides

Linear models and optimisation

Lectures 8–17

Lecture 08	Ensembles, weighted samples in trees, maths for ML, linear regression	Slides: Ensemble methodsSlides: Weighted samples in decision treesSlides: Maths for ML ISlides: Linear regression Notebook: Maths and JAXNotebook: Low-rank matricesStreamlit app (source)
Lecture 09	Linear regression, contour plots, geometric view	Slides: Linear regressionSlides: Contour plotsSlides: Geometric view of linear regression
Lecture 10	Linear regression II	Slides
Lecture 11	Gradient descent	SlidesReading: Taylor series and gradient descentReading: Gradient descent Video: Taylor series Notebook: Taylor series
Lecture 12	Gradient descent II	Slides Notebook: Gradient descent
Lecture 13	Gradient descent continued, ridge regression	Slides: Ridge regressionReading: SGD as an unbiased estimator Streamlit demo: Ridge regression
Lecture 14	Ridge regression and lasso	Slides: LassoReading: Interactive article on optimisation algorithms
Lecture 15	Logistic regression I	Slides Notebook: Logistic regression (best run locally for the interactive plots)
Lecture 16	Logistic regression II	Slides
Lecture 17	Logistic regression III	Slides

Neural networks

Lectures 18–23

Lecture 18	Multi-layer perceptrons I	Slides
Lecture 19	Multi-layer perceptrons II	Slides Notebook: Neural network
Lecture 20	Multi-layer perceptrons III	Slides
Lecture 21	Next-word prediction	Slides Notebook: Generating names with MLPs
Lecture 22	Convolutional neural networks	Slides: CNNsSlides: 1D CNNsReading: Equivariance vs invarianceReading: How CNNs work Notebook: CNNNotebook: CNN edge detectionNotebook: 1D CNNCNN Explainer
Lecture 23	Autograd and naive Bayes	Slides: AutogradSlides: Naive BayesReading: Chain rule Notebook: Autodiff

Naive Bayes, neighbours and clustering

Lectures 24–26

Lecture 24	Naive Bayes and k-nearest neighbours	Slides: Naive BayesSlides: KNN
Lecture 25	KNN, parametric and non-parametric models, recommendation	Notebook: Parametric vs non-parametricNotebook: Movie recommendation
Lecture 26	Curse of dimensionality and unsupervised learning	Reference: PCA (video) Notebook: Curse of dimensionalityNotebook: Image segmentation with k-meansNotebook: Image completionSegment Anything demoK-means visualisation 1K-means visualisation 2

Support vector machines

Lectures 27–29

Lecture 27	Constrained optimisation and support vector machines I	Slides: Support vector machines I Video: Constrained optimisation (self-study)
Lecture 28	Support vector machines II	Slides
Lecture 29	Support vector machines: soft margin	Slides

Lectures ran on Tuesdays and Thursdays from 3 January to 25 April 2023. Quiz dates are on the Assessment page.