Learn data science and more with new and improved MOOCs

Learn data science and more with new and improved MOOCs
Credit: Corbis

With recent enhancements, you can now get a truly useful education from Coursera and its ilk -- in data science, machine learning, and many other subjects

I’ve written about MOOCs in general and Coursera in particular in past articles, which resurface to the top of the Internet’s memory periodically. Whenever that happens, I try to reply to all the emails from readers, but sometimes I'm overwhelmed (email me again later and I’ll reply when I come up for air).

I realize why I'm in demand: I'm immensely qualified. According to Coursera’s statistics, I’m part of the impressive 95 percent failure rate for, like, all the courses I’ve taken.

That's because I rarely take the tests or submit the homework (though I often do the homework). I’m usually there to learn, so I don’t really care about being evaluated or getting a certificate. I’ll take the tests only if they're both hard and I have time for them. Thus, I come back to Coursera to fail course after course, which has made me a regular on the site and a close observer of how it's evolved.

Self-pacing at last

The most recent change at Coursera is the conversion of some of the courses to the online, learn-at-your-own-pace format, which is a very, very positive development. In fact, one of the best things to happen yet is that Stanford’s machine learning class is one of them. That’s right, enroll anytime you like, and proceed as your schedule allows.

Of course, you can succeed only if your schedule accommodates a few hours per week of study, but if you have a break in the middle for a conference or a work deadline, no biggie. Stanford is making some of its other comp-sci courses available in self-paced format, too.

More important, Coursera, Udemy, and Saylor have self-paced remedial math classes you’ll probably need to take in order to understand the entirety of the machine learning course. Particularly entertaining is Ohio State’s Mooculus. (Note: I recommend going to the OSU site for this course, because the OSU site seems to have been updated more recently -- and the Coursera doesn’t show you the book or the exercises.)

If you find yourself struggling with calculus, statistics, and machine learning because you were too focused on the cute girl or guy sitting next to you in math during high school or college, perhaps you should take Learning How to Learn now that you're less hormonal. This is one of my favorites, in part because the instructor, Barbara Oakley, is a badass who worked as a translator on a Soviet trawler and as a researcher in Antarctica.

Getting specialized

When I created a list of online classes you could take to get the kind of knowledge my company was looking for, a former professor who works with me noted it was actually more coursework than you’d normally need for a major.

Coursera has also created “specializations,” which are essentially tracks of courses in the same vein. The John Hopkins University Data Science specialization looks excellent. I’ve spoken to a few people who've taken it, and I’m enamored with their newly improved brains.

A downside is that specializations are still within specific university borders and composed of many courses that are not self-paced. I've wanted to take a number of courses on Coursera, only to find there was no current session scheduled -- or when it came up, I was too busy. This could easily happen if you invested in a specialization. What if there were an R language class from some other university? Why can’t I swap that in for the John Hopkins one?

What they need to do next

More self-paced courses is a positive sign, but the major MOOCs haven’t quite gotten it right yet. For example, the forums are now mostly abandoned wastelands. It reminds me of when you get an obscure error in Hadoop, so you turn to Stack Overflow and find someone else who has asked about the same error ... and has gotten crickets in response. That’s mostly what you’ll find in the forums of Mooculus, for instance.

Rather than paying $49 for a certificate of completion, which I could care less about, I'd prefer to pay those adjunct professors and grad students to stick around and answer questions -- and do the occasional WebEx if I’m really stumped. What if for a flat fee to the MOOC provider of, say, $149 every six months or so, I could get the full college experience (minus the entitled professors with tenure who don’t keep regular office hours)?

That modest fee would help the grad students and adjunct professors earn more than McWages. If only 1 percent of students in major subjects paid, these instructors would make a nice cut even after Coursera took its share. The courses would be self-paced to a point, but assisted if you need it.

Also, some form of swapping within specializations needs to be allowed. Crossing borders should be one of the key advantages of MOOCs that's difficult to replicate in a university. Certainly it doesn’t matter if I take statistics from John Hopkins or Duke.

Another observation: It seems that Coursera has deduplicated courses, but that's not good. Not only should there be the same course taught by different professors and/or different schools, but there should be ratings -- not merely stars, but specific ratings similar to those posted on RateMyProfessors.com, but aimed more at the course than the teacher. Sure, you can find ratings for some of the professors on Coursera (such as Fowler from Mooculus), but those should be integrated into Coursera and the rest. Maybe I want an “easy” course on R or maybe I want a hard one; either way, I want to know what I'm getting into. I should be able to choose using, like, data and stuff.

My thesis

MOOCs still get a bad rap, mainly from the entrenched or those expecting it to be “exactly the same” but online. MOOCs are still new and developing. Coursera, the monster of them all, is still changing and rapidly improving.

Overall, I’m happy with what I’m seeing. What's missing from MOOCs is personal attention. That isn’t out of reach -- and is probably an economically viable model for someone. As for standardized, crowd-sourced metrics on course quality, cooperation among colleges and sites will be necessary and will likely take a while.

MOOCs have a real shot at countering the way that higher education's ever increasing costs keep poor people out of the middle class. Free or nearly free online education is a revolution in the making, not only for the young, but for lifelong learning. At the very least, it's a great way for experienced professionals to break into data science, even if they forgot calculus.

From CIO: 8 Free Online Courses to Grow Your Tech Skills
View Comments
Join the discussion
Be the first to comment on this article. Our Commenting Policies