I was first acquainted with Continuing Professional Education (CPE) while working for a Big Six auditor in Detroit. Our CPAs were always worried about getting their CPE credits done on time. In those days, I was also working with some trade groups trying to create a certification program for computer professionals.
The Certified Data Processor (CDP) program never took off but, nonetheless, my profession became a hotbed of certifications for everything from tech support to Solutions Architect. Like other hiring managers, I take these “certs” seriously when recruiting.
My own tastes in CPE are rather eclectic. In this post, I’ll share some of my experiences so maybe if you’re new to this – or you’re an educator – it will give you some inspiration.
Superstar Lecturers on Video
I have written about Coursera before, so I won’t tell the whole MOOC story again, except to note that Coursera, Khan, and Udemy are now joined by edX, which is sponsored by Harvard and MIT so, you know – higher education will never be the same.
I have taken about a dozen courses through Coursera, including a “specialization” in deep neural networks. This one featured superstar AI educator Andrew Ng. As of this writing, Andrew is estimated to have taught eight million students.
This phenomenon is not new. Top professors often leave their colleges to go solo. I remember seeing Dr. Michael Hammer, inventor of Business Process Reengineering, lecture a packed conference hall in Washington, DC. As online education progresses, I predict the Pareto effect will set in, and schools will compete for a small number of superstar teachers.
It’s All About the Textbooks
Andrew’s classes were brilliant, but left precious little reference material. One lecture might have ten slides, and then you’re left to your own notes. So, concurrently with the Deep Learning course, I read Aurelien Geron’s book, Machine Learning with TensorFlow. In a way, it was good to have a book that was not the official textbook, because it gave me a different perspective.
I signed up for Stanford’s “Statistical Learning with Python,” strictly because of the textbook. It’s the classic from Hastie and Tibshirani (also superstar teachers) and it was just updated to use Python. I started going through the book on my own, and then discovered the class.
You can read a textbook on your own, but I don’t recommend it. I went through Barabasi’s book on Network Science, working exercises from each chapter in C# and Python. This was okay because I already had a background in graph theory. For the stats class, I often needed the lectures to clarify difficult points from the reading.
The lack of reference materials is especially frustrating if you take certification classes from vendors like Google and Amazon. You get a lot of advertising and a lot of video content, which is impossible to refer back to. Some of these classes are pretty chaotic, too, in terms of syllabus planning. Here again, I recommend buying a companion book.
Practical Coding Exercises
Coursera has a system where you can work Python notebooks inside their training environment. I think this is a clever way to keep their IP locked up, but I was always terrified it would time out and I’d have to start the exercise over again.
The edX class I’ve just completed simply leaves the code on GitHub. This is the most normal thing, from a developer’s perspective, and then students can use their favorite notebook. I copied all the labs into Google Drive and worked them in Google Collab.
Forums and Engagement
Discussion forums are a challenge. You’re never going to have the cohesion and topicality of a college course, because everyone is learning at their own pace. On the other hand, there are massive numbers of students encountering the material continuously, so this can be turned to advantage.
Stack Overflow is a popular online forum where you can find useful coding help – even if your question was last engaged ten years ago. In fact, many of Andrew’s coding assignments are discussed there. Over time, I believe an in-class forum could accumulate a critical mass of answers just like Stack Overflow.
On the other hand, expectations are different. If you’re paying for a class, and you’re stuck on the homework, you want an answer right now – especially if you suspect there’s a bug in one of the assignments. So, that’s down to the teaching assistant, or a staff of teaching assistants, or maybe ChatGPT.
Graded Tests and Certification
Since this is an ongoing hobby of mine, I looked into getting an online master’s degree, and decided against it for the reasons given above. If it’s just going to be slideware and a no-name teacher, I can do better on my own. What I’d really like to do is, go on picking best-in-class courses and somehow stitch them into a degree program.
Remember back in high school AP calculus when you learned about Lagrange multipliers? – Daniela Witten
Grading performance for online coursework is still an open problem. For professional certification, you take the class and then sit for an independently proctored exam. If you have the job experience, you may not need a class at all – and I think this hints at a solution.
The professional certification classes are not up to university standards, but people tolerate them because they need the cert. What the MOOCs could use is a more general (and recognized) system of testing. Instead of certifying that you can use a tool to do a job, like running Salesforce, there should be tests to show that you know something, like English Lit.
I am alluding, of course, to the College Board’s Advanced Placement (AP) program, which already offers college credit for passing their exams. This could become the certification regime for university aligned MOOCs like edX, and then they could organize degree programs around curricula supported by the exams.
