Wolfram Blog
Abrita Chakravarty

Designing the Wolfram U Data Science Course for Learning the Multiparadigm Workflow

August 15, 2019 — Abrita Chakravarty, Training and Development Specialist, Wolfram U

Designing the Wolfram U Data Science Course for Learning the Multiparadigm Workflow

A few weeks back, we announced Wolfram U’s latest open online course: Multiparadigm Data Science (MPDS). This course gives a hands-on introduction to basic concepts of data science through a multiparadigm approach—using various types of data, modern analytical techniques, automated machine learning and a range of interfaces for communicating your data science results. Our goal is to increase your understanding of data science while allowing you to take advantage of multiparadigm insights—whether you’re a newcomer working on a simple problem or an expert using well-established methods.

As the content creator and instructor, I’d like to provide some background on myself and my approach to the MPDS course. Beyond doing data science, I’ve found that multiparadigm principles make both teaching and learning more effective. In this post, I’ll give insight to the design of the course—the main goals, what topics are included and how to use the built-in interactivity to get the most out of your experience.

About Your Instructor

I work as a training and development specialist with the Wolfram U team at Wolfram Research. My academic background is in communications engineering and computer science, and I developed a passion for applied data science while studying computational genomics at graduate school.

My experience as a programming instructor at Duke University and as a Wolfram certified instructor has given me valuable insights on introducing simple concepts to students with the help of computation and programming. I have found the power and flexibility of the Wolfram Language extremely helpful in this context. In fact, my favorite part of creating the course was organizing the rich collection of my most useful Wolfram Language functions according to the MPDS narrative. These are functions commonly used by many in different scenarios in everyday programming tasks. When organized as building blocks for the MPDS workflow, the functions provide a more practical roadmap for navigating a data science project. So I really enjoyed designing this course, which demonstrates MPDS as best practice and provides a comprehensive look at the Wolfram technology stack that makes MPDS easy to implement.

Deciding What to Highlight

I personally believe it’s impossible to cover all topics in data science exhaustively in one course. Being a data scientist requires a broad combination of analytical skills and subject expertise that’s difficult to cover in a traditional instructor-led course. So rather than attempt to fit in everything, I wanted to give a starting point to explain the approach and get you exploring on your own. In this interactive course, I tried to break that extensive spread into bite-sized chunks.

Throughout the course, I keep the core idea of MPDS in focus: your workflow (and the resulting insights) should be driven by questions, rather than being confined by the standard techniques specific to the data or subject at hand. It is possible to utilize algorithms and techniques across disciplines like machine learning, statistics, signal processing, classical modeling, image processing, data visualization, traditional math, graphs/networks and more. Wolfram technology allows you to unify your development process across multiple subjects and paradigms, as well as lets you work with different types of data: flat files or databases, audio, images, sensor readings, text or arbitrary data scraped off the web. From that starting point, you can assemble a broad, flexible computational toolkit to integrate data processing, analysis and visualization capabilities into one start-to-finish workflow. You can learn more about the question-driven multiparadigm approach from this recent post or from the MPDS website.

Utilizing an Iterative Workflow

MPDS—and this course—can be viewed as an iterative process. In a data science project, the best way to make progress is to iterate through the stages of the workflow repeatedly, tweaking and improving each stage as you go. It is rather restrictive to design a complete project in one pass and give up on the opportunities to explore and experiment. Making multiple passes through the process provides the opportunity to add something new and useful—bringing up new ideas, opening doors to different kinds of analyses, incorporating more and different data sources, etc….

Similarly, this course starts with an overview and then goes on to revisit topics in more detail. I included a broad range of examples to cover different subjects and types of analysis. You can skim through the different sections and segments, go back and revisit specific parts of interest to you, work on quizzes and exercises to improve your understanding and explore further using the references provided.

Course Navigation

Titled Building a Project Workflow, the first section of the course highlights the usefulness of having a flexible, modular, iterative process for practicing MPDS. Using Twitter data analysis as an example project, this section introduces the stages of the project workflow: Question, Wrangle, Explore, Analyze and Communicate. Further sections of the course delve into specific parts of this process, experimenting with a variety of techniques that achieve the goals of a given stage.

Each lecture video moves quickly through the relevant concepts and functionality, with corresponding lecture notebooks containing all the code shown in the video. Course-takers can copy code into the Scratch Notebook pane at the bottom, immediately evaluate it, then edit and build on it for a deeper understanding of the topic. I find this layout especially helpful for this type of course, as it gives individuals the unique flexibility to learn in the way that makes sense to them—whether that’s by watching, reading or doing.

MPDS course layout

The quiz at the end of each section provides the opportunity to quickly review some of the functionality covered in the section. We’re in the process of also adding interactive exercises to accompany each lecture video. Successful completion of the exercises and submission of a project adhering to the MPDS workflow will earn advanced levels of certification for the course.

What Do You Think?

The MPDS course is our latest offering as a full interactive course, and we would love to hear back from you: what you liked, what you found especially handy or what was a stumbling block in your experience navigating the course. I hope you have as much fun taking the course as I had creating it. Happy explorations!

Take the interactive MPDS course now to streamline your data science workflow, or check out our Data Science & Statistics page for the latest Wolfram U events and courses.

Leave a Comment

One Comment


Manfred Plagmann

I went through the course as soon as it had been made available. As a well seasoned researcher and MMA user I had no interest in the certification but I enjoyed the course very much. I learnt some commands and functionality I hadn’t come across before. Using those has made some of my own analysis faster now. I have also taken some of your workflow on board to make my data analysis more robust. Thanks, Abrita, for the inspiration to review my own approach and improve on it. Please continue to creating these high quality course materials.

Posted by Manfred Plagmann    August 20, 2019 at 5:48 pm


Leave a comment in reply to Manfred Plagmann