UBC MDS

On April 6, 2019, EasyMarkit hosted their first Hackathon in Vancouver where teams were asked to offer an AI solution to improve patient communication. My team (Bailey Lei, Alex Pak, Betty Zhou) was awarded first place based on the accuracy of our model in predicting communication response from patients.

About the EasyMarkit AI Hackathon

EasyMarkit is a Vancouver-based company that focuses on developing automated patient communication software specializing in dentistry. The hackathon challenged participants to predict whether or not patients will respond to communications like appointment reminders. The event was hosted at EasyMarkit’s office in downtown Vancouver, and ran from 9am to 5pm. Owen Ingraham, CTO at EasyMarkit, and his team did a fantastic job of hosting the event and our team was extremely impressed by the culture at EasyMarkit.

When I first learned about convolutional neural networks (also known as CNNs, or convnets), I was shown a picture much like the one below, which is from the AlexNet paper:

I will call this a “row-of-boxes” diagram. In my experience, this type of diagram is common in both CNN papers and CNN lessons, even though the audiences are very different in those two contexts. I would argue that row-of-boxes diagrams are targeted at people who already understand CNNs, not students seeing them for the first time. Given the diagram issue and other obstacles, it took me a few iterations to fully understand CNNs when I was a student.

Since launching the UBC MDS program in 2016, we’ve received a lot of questions on why we designed MDS the way we did. The post will address the following design decisions:

• Statistics and CS as the home departments.
• Goal of the program: responsible use of DS.
• Length of the program: 10 months.
• Length of the courses: 4 weeks.
• Creating all new courses from scratch.
• The program prerequisites.
• Dividing the instructor role into two pieces: lecture and lab.
• Setting a single deadline for all weekly assignments.
• A few words on the curriculum.

Stat-CS partnership

The UBC MDS program is an equal partnership between the Department of Statistics and the Department of Computer Science. I have noticed that some other DS programs lack the statistics component, and I think that is a big loss. To us, data science is not just computer programming and machine learning (and certainly not just deep learning!). Rather, it is a broad field about how to ask and answer questions using data.

Last week we welcomed our third MDS Vancouver cohort. It only seems a short while ago when we launched the program with a cohort of “only” 22 students. Now we have 70 new students in the program with another 28 at our UBC Okanagan campus.

One of the strengths of the MDS Vancouver program is the diversity of our students:

• 70 new students.
• 41 domestic and 29 international students from Brazil, China, Hong Kong, Ivory Coast, India, Mexico, Puerto Rico, USA, Turkey and Pakistan.
• 32 women and 38 men (46% women, 54% men).
• a variety of academic backgrounds, from psychology, neuroscience, business, economics, political science, biology, chemistry, engineering and many more.
• Nine students have completed another advanced degree (Master’s, PhD, MD) prior to MDS.
• 60% of the class completed their Bachelor’s degree in 2016 or earlier, with 29% having just graduated in 2018.

We were extremely impressed with the quality and size of the applicant pool this year. The program is becoming more and more competitive but with more spaces opening up at our UBC Okanagan campus as well as a new MDS program with a specialization in Computational Linguistics coming in September 2019 (pending provincial government approval) there will be more opportunities for applicants to join our MDS family.

Before starting the Master of Data Science (MDS) program at UBC, I had been working as a civil engineer for a consulting firm in Vancouver. I soon realized that despite producing vast amounts of data, the civil engineering and construction industries have felt little influence from advancements in data science and machine learning. Having seen the capabilities of data science and machine learning implemented in other industries, it became apparent that the civil engineering and construction industries could benefit greatly through the implementation of these technologies.