Machine Learning Fundamentals

This is the second course from UCSanDiegoX (early 2018 run) that breaks the enrollment clause stating "Audit this course for free and have complete access to all the course material, activities, tests, and forums". There was no final exam for Audit students. Hopefully this will get corrected in the future runs of this course. The presence of staff on the Audit mode forum is very scarce, and the forums are divided into Verified/Audit students.

The course provides the "fundamentals" in only first few chapters. The more the courses advances the less mathematics is presented which hampers understanding of the subject. The quiz and homework problems have unlimited (!) number of attempts and provide no (!) explanations after the correct answer is given. This has been explained by the staff on the forum as a result of lack of time for preparing the course. If the missing explanations were provided, I would be giving the course 4 stars, and 5 stars if optional, more advanced mathematics videos were present.

The good things about the course: all materials were released upfront, which allows students for time management. There are many quiz and homework problems per week (order of 30) so there is an opportunity to practice, and the quality of jupyter notebooks is high, but both problems and notebooks are rather easy.

The instructor, despite shying away from math manages well to communicate intuitions about the methods used in machine learning, but Caltech's "Learning from data" https://work.caltech.edu/telecourse.html is still the best introduction to machine learning.

I have a few things to say about this course. Firstly, the lectures content is very good as the main concepts of ML models are explained well by Sanjoy Dasgupta. However, he misses on explaining the F1-score, precision and recall metrics which are often used in solving practical problems.

Secondly, the course does not teach any python code which is fair enough. however, it asks to do the programming assignments in python and these are quite challenging with no solutions provided. One of them for week 7 even asks in a form of a quiz question 'Were you able to correctly implement the kernel Perceptron algorithm?' giving the options 'of yes', 'almost' and 'no, I had a lot of trouble with this'. Unfortunately if you've chosen the third option, there is no feedback to help you.

What makes it worse that the course staff do not monitor the discussion forum and those stuck with these assignments are unable to get help with their questions.

There is no way to contact the course staff either, as opposed to other courses on Edx which display the teaching university contact information on the course introduction page.

To sum up, this course is brilliant for auditing (i.e. watching the lecture videos) but becoming a verified learner is not worth it due to lack of staff support.