After a week of everything going wrong, we are in a new week. So far everything has been going right. Picking back up the blog schedule, I have the latest Kaggle course I have completed. It is the Kaggle Intermediate Machine Learning Course.
The Good
Unfortunately, the good is not as good as the other courses. The courses are solid but very quick to read. The author of the course also provided knowledge on Pipelines which are freaking amazing. Learning about those makes the whole course well worth the couple of hours it takes.
Other than the pipeline, you can expect solid course that explain things well. Don’t expect much reinforcement. If just the new stuff is what you love then you will love this course part of the course.
One last positive is the data leakage was really good. The exercises not so much but I think the author did as well as they could with such an abstract topic.
Kaggle Intermediate Machine Learning Course Bad
Oh there was bad in this course. If you don’t have some background in this stuff, what they ask you to do would be next to impossible. The course swings from so easy that you wonder why they ask the question to so difficult that you are looking over all the course notes in the other courses to figure out what you should do.
This stupid easy to viciously difficult makes this one a much harder course to complete. In one lesson, new stuff was actually introduced in the middle of the lesson. Another, I had the correct answer in the fifth test that was graded incorrect because I got my previous example wrong, despite being given a correct on it. That one hurt.
Another example provided the same output as the desired answer and it should have been in the same directory type as well but was graded wrong. That one put me through a beating.
It is not all bad but the bad is in the exercises. You can battle through this if you have some programming knowledge.
Hopefully I don’t sound too negative on this course. While most of the exercises were not that great, the author nailed pipe and data leaking section. It is solid and the pipe exercise hit the right level too. There are just some glaring pain points. Don’t let this scare you away from doing it. It has its problems but the person teaching it has a solid grasp on the topic and explains it well.
I can’t agree more! I’m a full beginner, and for me, going through the Python exercises only relying on the tutorial given is next to impossible, without prior coding experience. Kaggle treats its exercise as more like another chance at indirect tutorial, introducing new concepts on the fly with the solution it gave. Which is really great to expand the student’s knowledge, but ONLY IF those new concepts also received some serious information along with it! A lot of times I’m stumped even with the solution open, as it involves shortcuts not yet explained / or a major jump in logic complexity from the tutorial it gave beforehand. And no explanation comes along with the solution, other than the comments. Comments are like saying things like #’this will do B’, but I have no idea how the function can actually do so. It’s not even in the help syntax. Only when I happen to come across the concept as a shortcut / or other usual field practice tips from other sources (around 3 months later) only then I get it. And I found two exercises where correct answers are deemed incorrect, with differences only in variable names. I even asked two IT professionals to see if there IS actually any differences, in case my eyes missed anything. None. Not even restarting works. Only by copying the solution the answer is deemed correct. Don’t get me wrong, I love the difficulty jump, as it sets itself apart from other online courses out there, but it’s detrimental for new learners without additional explanation for the solutions.