Great

Good!

good

good

GOOD

good

nice

good

good

GOOD

Good

Good

good

good

Good

Good

good

ok

wow

ok

Ok

.

PRO: Good overview and basic introduction of common machine learning techniques.

CON:

- The final assignment is peer reviewed! I saw no mention of this before purchasing the course. This means you are at the mercy of other students who may have less experience than you and may notbe qualified in grading assignments. Also it may mean you have to wait a long time before you get your certificate. It would be better to implement a Kaggle-style assessment of the models and use that to obtain a score and turn that into a grade. This would be transparent and instantaneous.

Some of the forum answers provided by the teaching staff are half baked and often inconsistent. e.g. they give example code for making a figure and and also a figure. But the figure is obviously not made with the provided code and the code contains typos. This is frustrating and makes learning harder than it should be.

Some of the code in the lab exercises don’t obey good practices. e.g. in every lab the data is normalised before train/test splitting. In the final project there is a comment that this should be done the other way around (and it really should!). Why not do it the right way in all the examples throughout the course?

The course provides a good overview of the topic over 5 weeks plus the project week. With previous knowledge of Python, the coding is easy to follow. The videos are good. However, the Python Jupyter notebooks provided could be significantly improved. The content could be of better quality and more rigorous. The notebooks have many spelling mistakes, few explanations, unnecessary imports, a few bits of code that are incorrect and need to be fixed, some unnecessary or incorrect statements, etc. Some of the exercises proposed in the notebooks are meaningless for learning. Better practice tasks could be thought. Different notebook parts are clearly written by different people with different coding styles, which can sometimes be confusing for the learner. The assessment (classifier of loan repayment data) could also be improved as it was confusing in terms of what data sets should be used for training and testing. Peer-review is perhaps not the best for assessment grading either. Overall I enjoyed it and learnt, it's a good first impression of the subject but I would have expected higher quality of the materials from IBM - Coursera. Also, it would be good if notes or slides were provided.

This course is pretty dense in mathematics concepts for evident reasons but there is a lot of repeat on "beyond the scope of this course" so maybe, the course should focus more on what they want to teach. This is the ninth course I took as part off the IBM professional curriculum and they all are formatted in the same way: the videos explains the concept in simple terms but you are left alone with the hands-on labs where you mostly learn nothing as you just execute the cells one by one until the end where you have to GUESS what should be written when the solution is not in the notebook. Now, during the whole length of this course, the labs are focused on how to create the different algorithms with an abrupt ending but the final submission leaves you having to come back to the week 3 trying to understand how to APPLY the models on another dataset. Given the global level of these courses which is supposed to be targetting beginners, I found the last submission to be harsh and from what I have read for the next course of the curriculum, the next one is even worse.