Which machine learning course on Coursera is the best?

Question

Machine learning

I am lost in the choice of which specialization to choose: from MIPT and Yandex ( https://www.coursera.org/specializations/machine-l... ) or from the University of Washington ( https://www.coursera.org/specializations/ machine-l... )? There is no time to master both. Are there any objective criteria in such matters?

Answer 1

The Washington specialization in ML seemed to me successful. Recently I took her first course - "Machine Learning Foundations: A Case Study Approach" , now I'm studying the second - "Machine Learning: Regression" . After the ML course from Andrew Ng, it seemed at first simple, but despite the more applied nature of the Washington specialization, it includes some algorithms that Eun did not have enough space for.
Also a big plus - at the end of each week there is not only a test, but also a large practical task (iPython Notebook, GraphLab, NumPy modules, etc. are used), and at first ready-made working tools are used, and as you progress, you will need your own algorithms write in Python. It can be seen that enough work has been invested in the course, and this is an indicator that the course will be useful.
If we take for comparison the Johns Hopkins University course "Practical Machine Learning" from the Data Science specialization , then there is less dynamic and not enough practical tasks.
There is another similar specialization from the University of Washington - "Tackle Real Data Challenges", but it is very difficult to perceive it due to the rather monotonous presentation of the presenter, plus a strange practice is proposed - it is proposed to participate in any contest on Kaggle (which in itself, of course, is a mandatory thing, but training tasks are also needed) and write a review on it, which is much it would be more suitable for a course of a humanitarian orientation, but not a technical one.
PS I have a skeptical attitude towards Russian-made courses. If the Americans do not pull in half the cases, then what can we expect from domestic manufacturers. With the specialization of MIPT and YandexI, of course, pre-familiarized. The beginning is dynamic - teachers do not seem to suffer from mumbling, they actively gesticulate (which is also very important for faster perception and concentration of attention), flowers in the background, beautiful furniture - all this is very good. If there is time between the Washington courses, I will definitely take this course as well.
So far, only the first course is available.- there is no actual machine learning (only linear algebra, the beginning of analysis and Python). But there is already a bad trend - the course consists of only 4 weeks (it is clear that you don’t want to spend more on mathematics and basic knowledge of the modules, but what prevents you from adding something more substantial, more complex), and in Washington the first and second courses ( I haven’t looked further yet) - 6-week (in my opinion, a smaller number of weeks in any course is clearly not enough, given that the first week in any course is an introductory one), and in the first course at a fast pace, but a review is given in some detail on the main basic topics and algorithms of ML, except perhaps for neural networks.
So there is no particular reason not to perceive the MIPT and Yandex course as something more than an addition to the main Coursera courses or designed for a very beginner audience.