| Category | Difficulty |
|---|---|
| HW | 4 |
| Exams | 5 |
Intro to ML is one of the most popular CS electives at CMU, as ML has been a very hot topic in the last few years. The class will cover a good variety of ML concepts, but does not go too much into detail since it is just an introductory class.
- Decision Trees
- k-Nearest Neighbors
- Perceptron
- Linear Regression
- Logistic Regression
- Neural Networks
- PAC Learning
- MLE/MAP
- Naive Bayes
- Markov Models
- Bayesian Networks
- Reinforcement Learning
The class is very well organized, and follows the approximate structure:
-
Lecture about [ML method]
- Hand calculations to derive an important theorem about ML method
- Algorithm
- Applications
-
Written and coding homework about topic
- Perform hand calculations on your own
- Implement algorithm in code
Homeworks are straightforward implementations of algorithms that are covered in class. Lecture slides and the lectures themselves go into algorithms in a great depth, so the homeworks should not be too bad. In addition, since the algorithms are often well-known in the ML field, you can find lots of support for them online as well. The handouts provided for the homework is detailed, so make sure to read that as well. Definitely make sure to write your own tests for your program if it is possible, since that can help you uncover bugs you can't find on Autolab.
As a trick, sometimes your program might take a while to run on your computer. If you run your program on the Andrew machines at CMU instead, you might find a considerable boost in your program speed.
There are often also short-answers for homeworks. They give you LaTeX templates to fill in, but you don't need to know that much LaTeX to work with it.
One of the trickiest parts of the class is that there is a wide breadth of material so it is hard to learn everything in depth. For exams, it is important to know all your equations and algorithms. Fortunately, you have a cheatsheet so you can put those on there!
In addition, you can also do other school's exams for extra practice on the topic.
The good news about ML is that it is a very widely taught course, so there are lots of good materials about it outside of CMU. The most notable source for ML education is Andrew Ng's CS229 course at Stanford. His course notes are great and his class has lots of great content.
You can find course materials here (this looks like an archive of the old Github, which is now deleted.).