Course logistics
What is this course about?
This course is an introduction to ML. We will assume that you have never seen ML before. However it is designed at the maturity of a 3rd year engineering undergraduate student. Here is a list of takeaways from this course.
- Introduction to capabilities of ML.
- Concepts in design and analysis of modern ML systems. - Failure Modes of ML.
- Statistical and Computational aspects of ML.
- What matters and perhaps more importantly, what doesn’t, in an ML system.
Expectations from the participants
- 100% attendance (absence need prior permission. other commitments, academic or otherwise, are not valid excuses)
- no deadline extensions
- no partial marking
- no make-up examsExam format
All exams will be closed book, in the classroom. In fact, the instructor may also use <instructors-favourite-chatbot> to generate questions (and answer keys) with numerical values based on your rollnumber.
The exam is designed to be challenging for everyone. It will be designed to help you understand what you don’t understand. Like machine learning systems, you are expected to demonstrate generalization on sample questions not in your training data. Expect long comprehension style questions with multiple parts.
You are allowed 1 white A4 page (2 sides) of handwritten (in blue ink) formula-sheet.
Reference materials
As you may be aware, ML is a fast moving field. The field is moving at a break-neck pace and requires us to keep up with the developments. Many introductory textbooks have been written for ML. While there are some standard textbooks that respect chronology in their development in the field, we will choose a slightly different approach where we build on concepts as we proceed into the semester. This means that we may cover some topics that were only discovered in 2020 far before algorithms invented in 1990.
There have been some significant findings in the past decade that has required rewriting textbooks. Hence we will follow an order different from most textbooks. We will build on things as we cover them.
That said, we will borrow material from several textbooks. There is no standard textbook that we follow. Here is a list of some intro to ML textbooks that I like (in no particular order) - Learning theory from first principles (Bach) - Foundations of Machine Learning (Rostamizadeh, Talwalkar, Mohri) - Pattern Recognition and Machine Learning (Bishop) - Elements of statistical learning (Friedman, Hastie, Tibshirani) - Understanding machine learning (Shai-Shalev, Schwartz)
Grading scheme
| Total | 100 |
|---|---|
| Class participation | 20 |
| Homeworks (4) | 10 |
| Quiz 1 | 10 |
| Midsem | 20 |
| Quiz 2 | 10 |
| Endsem | 30 |
Important dates for homeworks and quizes
| Week\(^*\) | Date | Event |
|---|---|---|
| 2 F | Homeworks 1 upload | Jan 16, 2026 |
| 4 W | Quiz 1 | Jan 28, 2026 |
| 6 F | Homeworks 2 upload | Feb 13, 2026 |
| 8 F | Midsem\(^{**}\) | Feb 27, 2026 |
| 10 F | Homeworks 3 upload | Mar 13, 2026 |
| 12 W | Quiz 2 | Mar 25, 2026 |
| 14 F | Homeworks 4 upload | Apr 10, 2026 |
| 16 F | Endsem\(^{**}\) | Apr 24, 2026 |
\(^*\) classes are on Wednesdays and Fridays. Homeworks are due in the next exam at the exam venue
\(^{**}\) these are my best estimates, these are decided centrally by the academic section