UBC MDS

When people stress the importance of good communication in data science, they are usually saying something like what Hadley Wickham says in his book R for Data Science: “[It] doesn’t matter how great your analysis is unless you can explain it to others: you need to communicate your results.” In other words, good communication is crucial to data science because it is the final step, without which all prior work would be wasted.

Over the years I’ve struggled with the disconnect between “algorithms” — as a student might see in a standard algorithms and data structures class — and optimization. Several of the algorithms taught in such courses are in fact instances of (discrete) optimization: for example, dynamic programming (DP), or Dijkstra’s algorithm for finding the shortest path in a graph (related to DP), or greedy algorithms for solving the traveling salesman problem. I like the example of dynamic programming in particular because it can often form a conceptual island in students’ minds, without apparent connection to any of their other knowledge (for example, in the famous CLRS algorithms book, DP is listed under the miscellaneous heading “Advanced Design and Analysis Techniques”). In this post I’ll do a deep dive into one of our MDS assignments (or “labs” as we call them — but they’re really assignments) that attempts to connect these concepts using a single application. The lab is available here, as part of our publicly available teaching materials.

This is an exciting time for the UBC MDS program, with the second cohort starting their journey in just a few days. We are thrilled about the strength and diversity of our incoming class. Here are some facts about the incoming cohort:

• 43 new students.
• 21 domestic, 22 international.
• 20 women and 23 men (47% women, 53% men).
• a variety of academic backgrounds, from psychology to civil engineering to a current UBC medical student who is taking a year out to pursue MDS.
• previous degrees from UBC, other Canadian universities, the U.S., China, Colombia, France, India, South Africa, Turkey, and Vietnam (see map below).

In this post I will describe the system we use for delivering courses via GitHub in the UBC MDS program. I will first describe our high level goals and how we tackled them, and then discuss some implementation details. This post assumes the reader is familiar with git and GitHub and focuses more on the technical setup than the pedagogical implications.

Our main goals for this project were as follows:

• help students gain experience with version control systems
• eliminate all paper submissions
• facilitate peer review amongst the students
• facilitate group projects/assignments
• allow course staff to view and comment on work-in-progress remotely
• have a single place where most/all course content lives

When we were starting to think this through (around May 2016) there were already several existing solutions, such as GitHub Classroom, but none met all of our needs. We were committed to using GitHub rather than GitLab/Stash/etc because one of our team members, Jenny Bryan, had already been using GitHub to deliver STAT 545 and we wanted to build on her experience. The general approach is that students submit their work via GitHub and the TAs view/grade it there as well.

Warning signs and tools to stop the spread of misinformation.

Have you ever seen an over-enthusiastic advertisement and wondered how much of it was true? If so, you’re already one step ahead of 67% of people! Okay… that was completely made up. But hopefully, you can see how easy it can be to trust numbers as facts, especially when they are embedded in nice compliments. With a world of information right at our fingertips, it has become increasingly difficult to decipher what’s true, what’s false, and what’s true with a little twist.