A very challenging course.

good

Pros:

Doesn't rush over the various methods, discusses them in detail. The lab videos are some of the most in depth in this specialization.

Discusses the logic of designing a good recommender system quite well.

Cons:

The content is out of date in applications. Like it seems every course in this spec, the tasks we are assigned and the tasks discussed are totally mismatched a lot of the time.

The quality of the course got worse towards the end: to much speach with few content in recommender systems, many unfcuntional notebooks.

The Qwiklabs materials are outdated, however the lectures are insightful.

some labs are deprecated.

Outdated (TF, GCP)

The first part on content-based systems was pretty good, but everything after that was a mess. The second instructor (Ryan) was talking way too fast, and it felt like he was rushing everything he was explaining, and it would've been alright if his explanations were, at least, comprehensive enough. The problem is, there wasn't much groundwork in the course to build a good foundation for the students; they just did a brief introduction to the concepts, then rushed through the code implementation. Slowing down the videos did not help at all; it actually made it worse. The labs are another story, but then everyone else seems to be complaining about that, as well.

This is the only course in both specializations (ML on GCP and Advanced ML) that I didn't like.

while the topics and lectures are very interesting, the course is extremely broken in its current form. There are multiple instances where you first get a lecture and are then asked to do a lab, but the lab is on a completely different topic than the lecture was. One example is the final lab of week 2, where you should be building an end to end solution, but instead you get a lab that only focusses on a topic from week 1. It seems this course was reworked at some point, but the 2nd version is no longer coherent whatsoever. It's hard to recommend this course in its current form as a result.

The organization and content were unclear and repetitive. The reinforcement section was tedious with too much theory and lacked practical application. The labs using TensorFlow were not suitable for beginners. Improvement is needed for a better learning experience.

My first one-star course on coursera which is especially disappointing given that it is a course from google. It is also a pity since the covered topics and approaches are very interesting. The main problems are: simpler methods are overcomplicated (first week) and more complicated approaches (week 2) are severely lacking math (especially RL). But the worst thing is the labs and how they are organized. They are simply unpleasant to do, and the code there is UNREADABLE and boring.