Foundations of Machine Learning

From first principles to a model you can defend

Dr. Ayesha Rehman

Professor of Machine Learning · Dean of the Academy

About this course

Machine learning is a discipline of judgement disguised as a discipline of code. This course builds the judgement first: what it means for a model to generalise, why validation is the whole game, and how the classical algorithms encode different assumptions about the world.

You will implement linear and logistic regression from scratch, train tree ensembles on a real Pakistani public-health dataset, and finish with an error analysis you would be comfortable presenting to a sceptical stakeholder.

Every module ends with a graded quiz or a lab that runs in your browser — no environment setup required.

What you will be able to do

  • Frame a business or research question as a learning problem with a measurable objective.
  • Implement gradient descent and understand what loss functions encode.
  • Choose and justify validation strategies that reflect how the model will be used.
  • Diagnose underfitting, overfitting and leakage with error analysis.

Curriculum

13 lessons · 3h 25m

  1. 01

    Learning from data

    What a model is, what it is not, and why generalisation is the only thing that matters.

  2. 02

    Linear models and optimisation

    Build regression from scratch so that every later model feels familiar.

    • Linear regression by hand18 min
    • Lab · Implement gradient descent35 min
    • Logistic regression and decision boundaries16 min
  3. 03

    Validation and error analysis

    The part that separates a demo from a deployment.

    • Cross-validation done honestly17 min
    • Metrics that match the decision14 min
    • Module quiz · Validation8 min
  4. 04

    Trees, ensembles and a real project

    The workhorses of tabular machine learning, and a project that ties everything together.

    • Decision trees and random forests16 min
    • Gradient boosting in practice15 min
    • Project brief · Maternal health risk20 min
    • Final assessment12 min