Recent Posts

I mentioned to a friend that I was learning Julia for a project, and he asked: “Why?”. The answer is that I need to use a Julia only library. However, there are some nice features that I’d like to share. My primary languages are Python and R so although I know some of these features are available in other languages they were surprising to me. If you are looking for a full tutorial, I would suggest JuliaBox which gives out free Jupyter compute for Julia learners.

CONTINUE READING

After eight months of coursework, the UBC Master of Data Science (MDS) concludes with a 2-month Capstone project. Our group’s project focused on analyzing data from UBC’s massive open online courses (MOOCs). UBC offers dozens of MOOCs to thousands of students through the edX platform. Unlike in-person classes, in a MOOC the instructor cannot observe student engagement directly. Instead, they must infer student engagement from recorded events. An example of course event would be a user pausing a video.

CONTINUE READING

If the No Free Lunch theorem says that all algorithms preform the same across all data sets, then why does XGBoost win so many Kaggle competitions?

CONTINUE READING

Recent & Upcoming Talks

More Talks

A First Glance at Medical Computer Vision Neural Networks with R and Keras
Feb 12, 2019 1:00 PM
Learn Data Science by Doing Kaggle Competitions: Fraud Detection
May 31, 2018 6:00 PM
The Automated Data Scientist: Tools for Feature and Model Discovery
Apr 25, 2018 5:00 PM
Intro to Deep Learning with Fast.AI: Recurrent Neural Networks
Sep 28, 2017 6:00 PM

Projects

Hyperopt-Sklearn

This is a fork of the scikit-learn wrapper for hyperopt, a hyperparameter optimization library. This fork is Python 3 compliant and contains more cross-validation options.

Pommerman

Pommerman is stylistically similar to Bomberman, the famous game from Hudson Soft. I am currently in a team competing to create the best multi-agent reinforcement learning models for Pommerman. I will release code after the competition ends.

Vizit: Visualization for Instructional Teams

My team and I built a dashboard that is accessible to MOOC instructors directly in their edX course website. The dashboard will help instructors to discover patterns across the structures of their courses, with a special focus on course elements where student engagement is especially high or low.

Recent Publications

. Identification of Sample Comparability Issues During the Iterative Design of Game-Based Cognitive Assessments. LAK 2019, 2018.

. Constructing Cognitive Profiles for Simulation-Based Hiring Assessments. EDM 2018, 2018.

. The Rise of Immersive Cognitive Assessments: Towards Simulation-Based Assessment for Evaluating Applicants. AIED2018, 2018.

PDF

Contact

I am available as a data science consultant. Please send your project proposal to the email below.