An Introduction to Machine Learning for Physicists
A five-lecture short course taught at the 6th Dr. Riaz ud Din Memorial School on Physics, University of Agriculture Faisalabad, August 2026
About the Course
I taught An Introduction to Machine Learning for Physicists as a five-lecture short course at the 6th Dr. Riazuddin Memorial School on Physics, held August 3–7, 2026 at the Department of Physics, University of Agriculture Faisalabad, in hybrid mode.
The course builds machine learning from first principles for an audience of physicists, assuming no prior background. Rather than surveying methods, it follows a single running dataset (a cricket match) from a line fitted by hand through to neural networks, then turns to real physics data: planetary orbits, where symbolic regression rediscovers Kepler’s third law, and the Iris dataset for classification. The final lecture covers generative models and language models.
Each lecture has worked examples, hands-on activities, and an accompanying Jupyter notebook. The complete notes include five appendices with supplementary derivations and a glossary organised by lecture.
Lectures
| # | Lecture | Topic | Materials |
|---|---|---|---|
| 1 | What Is Machine Learning, Really? | Fitting a line by hand; naming model, loss, and training | Slides · Notebook · Video |
| 2 | How Machines Learn | Gradient descent, overfitting, and regularization | Slides · Notebook · Video |
| 3 | Neural Networks | The artificial neuron, layers, and backpropagation | Slides · Notebook · Video |
| 4 | AI That Discovers Physics | Symbolic regression (rediscovering Kepler’s third law) and classification | Slides · Notebook · Video |
| 5 | The Bigger Picture | Generative models, language models, honest limitations, and a capstone project | Slides · Notebook · Video |
Materials
Complete lecture notes (PDF): all five lectures in a single document, with appendices on least-squares derivation, convergence, polynomial fitting, and network training, plus a glossary and a list of data sources and further reading.
Opening slides (PDF): course introduction and overview.
Recordings: all five lectures were streamed live and remain available on the school’s YouTube channel. Individual links are in the table above.
rmsop-2026-lecture-notes: full LaTeX source, notebooks, figure-generation scripts, and data. Figures are pre-generated and committed; each lecture folder has a make_figures.py to regenerate them.