Some months ago I started the Microsoft Professional Program for artificial intelligence. Getting the certificate cost me 800$ and about 48 hours of time.
Before you consider investing that much money and time lets have a look at what you get.
What is the Microsoft Professional Programm?
The Microsoft Professional Programm is a cooperation of edx.org and Microsoft. It is basically an online video course with automated tests to control what you have learned. After finishing the course you are handed a certificate from Microsoft itself and not from the Partner edx.org.
The Microsoft Professional Programm is not to be confused with the MSCE and MSCD Programm. These are completely separate certificates.
How much time and money do I have to invest?
The courses are self-paced. If you are familiar with a topic you can complete one in 2 hours or less. If you have to do the learning it can take up to 10 hours. I didn’t need more for any of the courses. Even the ones I was really unfamiliar with.
Please check the end date of a course before ordering it! The courses have a 3 month time window to be completed and then restart. Even a paid certificate is ended when the course ends and you need to reapply if you haven’t finished it. And yes, you can enroll just a few days before course end which might be impossible to accomplish.
The capstone project(the final assessment) is a one month project that can only be done every 4 months.
Each course can be done for free, but the needed certificate is 99$ each. So you end up with roughly 1000$ until its a promotion (which it was for me).
The courses were well prepared and every tutor seemed to know what he was talking about. Tone and Video quality were good. The transcript helped a lot and I still keep some of the examples. Also, the statistics part was very helpful even if painful. The certificate makes a good impression on a resume. I learned a lot doing these courses. Even the bad ones.
Some of the tutors were clearly no teachers. They spoke to fast and furiously clicked through pages and pages of formulas and concepts. One was not understandable at all due to a thick accent. Many topics were therefore much harder than they have to be. Also, some tests were just not well prepared as they asked for topics that were not taught in the course or the questions were misleading.
The Microsoft Professional Programm Dashboard. This thing is embarrassing. It doesn’t update on time, it loses progress, leads to wrong courses and is a general pain. I see no reason a simple web tool like this is working this crappy. As I’m writing this all my progress for the data science track is gone and I have to contact the support.
Would you recommend it to get into Machine Learning?
No, I wouldn’t. I would recommend it for people who are already actively implementing intelligent solutions to have a certificate to show. In the state, it was when I did it the quality of teaching wasn’t on par with the quality of information delivered and it was much harder than it had to be. There are several better introductions around.
To make that clear: I learned a LOT from these courses, even the bad ones. But if this had been my first contact with Machine Learning I wouldn’t be interested in the topic anymore.
What courses are to accomplish and how is the quality?
On several occasions, you can choose courses to do. So I can only describe the courses I did. Some of them were replaced. The quality was very mixed.
1. Introduction to AI
This is the basic introduction course. It is very easy and I skipped most of it because the principles were known to me. Compared to the other courses the quality was ok and it is a good measurement if you should proceed. If this is hard for you, you might need more experience because the other courses are much harder.
2. Introduction to Python for Data Science
This course was, as many, overhauled after I did it. The quality is hopefully better as the whole concept wasn’t very practical and without prior python knowledge, it was basically not doable.
3. Essential Mathematics for Artificial Intelligence
A good mathematical introduction, but hard if you haven’t done math for a long time. The statistics part was especially excruciating but at least understandable with some effort. I was very glad when I was through.
4. Ethics and Law in Data and Analytics
This was the course I planned on skipping as fast as possible, but it was very interesting and I learned a lot. It never occurred to me that an ML model could inherit the bias from its training data and therefore could predict in a biased way.
5. Data Science Essentials
Another overhauled course. This one was interesting but didn’t cover anything that wasn’t also in another course. Good thing it was taken offline and replaced.
6. Principles of Machine Learning
This one was pretty ok I guess. It explains the whole Azure ML environment which enables you to test ML models fast and without programming knowledge. Some of it was already covered in Data science essentials so getting into it wasn’t hard. I still have some of the examples which I found promising. It is now replaced by “Principles of Machine Learning: Python edition” Which I’m doing at the moment.
7. Deep Learning explained
I really had to check if I have actually done this one and I have. It must have been unremarkable as I can’t remember a thing. The topics were known to me and it’s not one of the hard ones.
8. Reinforcement Learning Explained
Not a good course. Seriously. It is the worst one of them all. Reinforcement Learning is the most fun part of ML and they made it boring and overwhelming. This course is about the functions. The tutors were reading formulas for a long time and the examples were cryptic at best. Yes, all of the topics covered are important, but covering them out of context is ruining something that can be incredibly fun to learn. For a better introduction to the topic click here
9. Natural Language Processing
I’m sorry to say but this course was really bad too. It was very theoretical and the worst: I didn’t understand a word of what the tutor was saying. If it wasn’t for the transcript I would have given up. Also, the test questions often left the topics that were covered and once even asked for technical terms I couldn’t find anywhere. Neither in the course nor on google.
The capstone project was simple. You got a dataset of MRT scans and make a model predict which way it is turned(patient orientation). If you’ve done the MNIST before this was basically a one hour project. This project alwazs changes so I might have had luck.