HomeMachine Learning

Create an account to study the free first module. Start free.

Machine Learning

Field guide to the BITS ZG565
Machine Learning course.

Each session opens with a from-scratch ramp (concrete problem → interactive demo → plain-English math), then a formal treatment, cheatsheet, and practice — smooth to hard, not the other way around. Citations point back to Mitchell (Source), Bishop (Source), Tan/Steinbach/Kumar (Source), and Burges (Source).

15 sessions · 5 thematic tracks · 2 sub-routes each · cite-checks back to textbooks

Not started
Orientation

What is this course?

October 2026 exam format: closed-book midsem; open-book comprehensive (final). Check your cohort’s latest instructor instructions for permitted materials.

Machine learning is what you reach for when the rules are too long to write by hand. The course builds a small number of families of models — linear, tree, distance-based, probabilistic, ensemble — and asks you to understand each well enough to implement it on toy data, diagnose its bias-variance behaviour, and choose the right evaluation metric.

The course is layered on top of the math foundation: gradient descent, regularisation, and the matrix identities that power logistic regression all live a few modules back. The first sessions are deliberately gentle; the mid-course review (S08) is the pivot where everything starts connecting.

What you walk away with is a working mental model of when each family wins, plus a habit of asking four questions before fitting anything: where does the experience come from, what is the target function, what is the representation, and what is the search algorithm.

Mental model

Five nouns, one pipeline

Every session of this course is a different filling of the same five slots. Different models, different losses — but the pipeline is identical.

input x
one training example
model f
the prediction model
target y
the true label
loss L
how wrong the prediction is
optimizer
how to update the model
Scope

What this course owns — and what it doesn't

Owns

  • Implementing regression and classification from scratch on toy data
  • Closed-form least squares and gradient descent (batch / stochastic / mini-batch)
  • Logistic regression, log-loss, multi-class via softmax
  • Decision tree construction (entropy / information gain), pruning intuition
  • k-Nearest Neighbours, locally weighted regression, radial basis functions
  • Bayesian reasoning: MLE, MAP, Naïve Bayes, generative classifiers
  • SVM intuition: margins, kernel trick, support vectors
  • Ensembles: bagging, random forests, AdaBoost, gradient boosting
  • K-Means and the EM intuition for soft clustering
  • Evaluation: bias-variance decomposition, cross-validation, comparison metrics
  • Bias, fairness, and interpretability questions for deployed models

Doesn't own

  • Production MLOps — model serving, monitoring, CI/CD for ML
  • Deep learning internals — backprop through arbitrary graphs, transformer architectures
  • Reinforcement learning — policy gradients, Q-learning, actor-critic
  • Time-series forecasting — ARIMA, state-space models, recurrent architectures
  • Computer-vision or NLP pipelines from scratch — see BITS ZG565 electives
  • Hyperparameter search at scale — Bayesian optimisation, distributed sweeps
  • Causal inference and counterfactual reasoning — separate course
  • Probabilistic programming languages — PyMC, Stan
All 15 sessions

Pick a session

Open any session to get the from-scratch ramp, formal notes, cheatsheet, and practice set. Track groups below show the recommended order — S08 is the mid-course pivot, S15 is where everything comes together for evaluation.

Index

All 15 sessions, flat

Same sessions, just ungrouped — for skim-review or to jump to a specific number.

S01
18 min read

Introduction

What is ML, why it matters, taxonomy, designing a learning system, issues in ML.

Source: Tom Mitchell — Machine Learning. Chapter 1
Open session →
S02
22 min read

Machine Learning Workflow

Role of data, preprocessing, sampling, training, metrics for testing.

Source: Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 2, Ch3
Open session →
S03
18 min read

Linear Models for Regression

Closed-form solution, gradient descent (batch / stochastic / mini-batch), basis functions.

Source: Christopher Bishop — Pattern Recognition and Machine Learning. Chapter 3
Open session →
S04
18 min read

Bias-Variance & Intro to Classification

Bias-variance decomposition, discriminant functions, decision theory, intro to logistic regression.

Source: Christopher Bishop — Pattern Recognition and Machine Learning. Chapter 3, Ch4
Open session →
S05
18 min read

Logistic Regression

Logloss, gradient descent, multi-class logistic regression.

Source: Christopher Bishop — Pattern Recognition and Machine Learning. Chapter 4, Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4
Open session →
S06
18 min read

Decision Trees

Information theory, entropy-based construction, overfitting, MDL pruning, continuous & missing attributes.

Source: Tom Mitchell — Machine Learning. Chapter 3, Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4
Open session →
S07
18 min read

Instance-based Learning

k-Nearest Neighbours, locally weighted regression, radial basis functions.

Source: Tom Mitchell — Machine Learning. Chapter 8
Open session →
S08
18 min read

Mid-course Review (Sessions 1-7)

Map of sessions 1-7, five big ideas, formula sheet, decision-boundary sketches, book/resource map.

Source: Tom Mitchell — Machine Learning. Chapter 1, Ch3, Ch8; Christopher Bishop — Pattern Recognition and Machine Learning. Chapter 3, Ch4; Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 2, Ch3
Open session →
S09
18 min read

Support Vector Machines

Linearly / non-linearly separable data, kernel trick (Mercer), structured & unstructured data.

Source: Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4, Christopher Burges — A Tutorial on Support Vector Machines for Pattern Recognition. See the cited chapter in your copy.
Open session →
S10
18 min read

Bayesian Learning (Foundations)

MLE, MAP, Bayes rule, optimal Bayes classifier.

Source: Tom Mitchell — Machine Learning. Chapter 6, Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4
Open session →
S11
18 min read

Bayesian Learning (Naïve Bayes & Generative)

Naïve Bayes classifier, probabilistic generative classifiers, Bayesian interpretation of linear regression.

Source: Tom Mitchell — Machine Learning. Chapter 6, Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4, Christopher Bishop — Pattern Recognition and Machine Learning. Chapter 4
Open session →
S12
18 min read

Ensemble Learning (Bagging & Random Forest)

Combining classifiers, bagging, random forests, boosting intro.

Source: Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4
Open session →
S13
18 min read

Ensemble Learning (Boosting)

AdaBoost, gradient boosting, XGBoost.

Source: Tan, Steinbach and Kumar — Introduction to Data Mining. Chapter 4, Lecture Notes
Open session →
S14
18 min read

Unsupervised Learning

K-Means clustering, EM review, mixture models for soft clustering, applications.

Source: Tom Mitchell — Machine Learning. Chapter 6
Open session →
S15
18 min read

ML Model Evaluation & Comparison

Comparing ML models, bias, fairness, interpretability.

Source: Tom Mitchell — Machine Learning. Chapter 5, Lecture Notes
Open session →

Printing is disabled

For course navigation and interactive practice, please use the live app. Offline PDF generation is disabled.