Our interactive Multiparadigm Data Science (MPDS) course has been up at Wolfram U for over a month now, and we’re pretty satisfied with the results so far. Hundreds of people have started the course—including students from our first Data Science Boot Camp, who joined us at Wolfram headquarters for a three-week training camp. Thanks to the success of the boot camp, we have also had several projects submitted for advanced MPDS certification, which will soon be available within the interactive course.

But what exactly does it mean to be a practitioner of MPDS? And how might the multiparadigm approach improve my computational projects? To find out, I decided to try this free course for myself.

About the Course‐Taker

My background is pretty broad—I’ve done some technical support, audio engineering and web development, and my BS is in computer science and physics. Though I’ve had some experience with error propagation, linear regression and plenty of programming projects related to data analysis, I have never thought of myself as a data scientist.

As a technical writer for Wolfram, I get to play around with Wolfram Language functions here and there. I have had the pleasure of working on some fascinating pieces that really got my gears turning (text analytics and geovisualization are currently my favorite coding areas). And recently, I’ve become pretty well acquainted with MPDS and how our technology enables it. The approach resonates with me, and I’ve been thinking about getting more into data science.

I don’t normally have the right schedule for traditional classes or guided webinars. This open, interactive course seemed just right for my situation and learning goals. And it’s free, so of course I jumped at the chance.

Starting with a Concrete Plan

I’ve seen and used a lot of data science functionality before, but mainly in isolation. Getting quick, high-level output is easy with the Wolfram Language; for me, the challenge is usually figuring out what to do next. So I appreciated having the full workflow laid out up front:

Mapping the stages out this way helps me better understand which functions are useful for which steps. And having a repeatable process makes data science seem more like something I can achieve. Sometimes starting with the right question and following a consistent process can do more to solve a problem than an entire collection of neural nets.

At the same time, the opening section emphasizes that the process is iterative: later stages can generate new information to feed back into the earlier ones. Though continually “starting over” might seem counterintuitive, I see this process as comparable to the Agile approach I was taught for software development. Besides, scientific discovery comes from trying a lot of different things. Why should data science be any different?

Course Highlights

This was my first time taking an online course, so I was glad to see both a transcript and an interactive scratch notebook. I think I finished much faster (and probably learned more) because those made it easy to follow along. As a relentless tinkerer, I found the scratch notebook extra helpful in the more code-heavy sections, because it allowed me to experiment until I understood exactly what each piece of code was doing:

The variety of examples throughout the course was useful for me as well. Seeing MPDS applied to a range of subjects gave me a lot of food for thought. Beyond that, it helped solidify the idea that methods and techniques don’t need to be subject-specific. For instance, while I’m familiar with using regression analysis as part of a lab experiment, I had never considered how it might apply to estimating a credit score:

Though it was by far the longest section in the course, the fourth section, Assemble a Multiparadigm Toolkit, turned out to be the most informative part for me. As Abrita’s post points out, there are a lot of different tools available, and this gave me a great overview without getting bogged down in unnecessary detail. I especially like that these later videos refer back to the Question stage, pointing out the different questions that might be answered by each technique described. That led to a few big “Ah-ha!” moments for me.

I also got a lot out of the final section, Getting the Message Across, which has some excellent examples of easy Wolfram Language deployments. Report generation is one functional area I’ve always been curious to explore but have never managed to get the hang of on my own. The examples here worked well as springboards for creating my own reports:

The quiz questions are mostly about Wolfram Language functionality, so I found them pretty straightforward. Even for those less familiar with the language, the videos answer the questions directly. And if you’re having difficulty, you can always review the content and try again. Once I completed all the videos and quizzes, I earned a certificate:

Hungry for More

After finishing the course, I walked away with a lot of questions—but in a good way! I found myself thinking about how I could apply my new understanding to my own computational ideas. How can I represent my data better? What happens if I swap classification methods? Will a different visualization show me new patterns?

That questioning nature is what drives successful exploration and discovery, and it’s a big part of what makes MPDS so effective. Following the multiparadigm workflow can open up new avenues for discovering all kinds of unique insights. But don’t take my word for it—try it for yourself!