jason.n.deon [at] gmail.com

Research

Since 2019, I have been pursuing a PhD in Computer Science at Dalhousie University under Dr. Sageev Oore. I am also affiliated with Vector Institute. My most recent research projects involve using machine learning to create new, innovative tools to produce music. More broadly, my research is focused on cutting-edge sequence models for prediction and generation. In addition, I am interested in applications of representation learning for generative models.
Before Dalhousie, I completed an MMath degree in Pure Mathematics at the University of Waterloo, where I studied Riemannian geometry and its applications to machine learning. I also have a BSc in Mathematics and Computer Science from Saint Mary's University, during which I was involved in a wide range of research labs, studying topics such as machine learning, data science, graph theory, and nanochemistry.

Piano and Music Production

I am a musician hobbyist, starting way back in piano lessons in third grade. I have a great love for the piano, specifically this one:
In most recent years, I tend to play classical pieces (my favourite composers are Chopin and Rachmaninoff), but I also occasionally play jazz or music from the soundtracks of video games or movies. Sometimes I put on recitals for family and friends or do personal recordings. Below are two excerpts I recorded, which are written by Rachmaninoff and Vince Guaraldi respectively.
Moment Musicaux No. 5
Happiness Is
When I am in the mood, I will sometimes write my own music, mostly inspired by soundtracks from video games I have played. I typically work with MIDI and soundfonts as opposed to recording actual acoustic or electric instruments. Here are three samples of original tunes created with MIDI using various restrictions: the first uses a variety of free public domain soundfonts; the second uses soundfonts from a SNES game, Secret of Mana; the third uses synths emulating a Game Boy soundchip.
Bossa Nova
Secret of Mana
8-bit Theme

Recreational Math

I use math in much of the work I do, but I also get great pleasure from diving into the vastness that is pure mathematics (the study of math without intending to apply it practically). One of the websites that also does a good job of reminding me what I love about math is Project Euler, which contains a large repository of challenging math problems, typically tackled with programming. As of 2022, there are now over 800 archived problems, of which I have solved over 200. My personal favourite problems in terms of cleverness and elegance include #148, which uncovers some beautiful patterns in Pascal's Triangle, and #202, where the right perspective turns a continuous problem into a discrete one.

Unpublished Works

Here are a list of writings I have worked on that are not otherwise shared on the internet.

Master's Research Paper

Applications of Information Geometry to Machine Learning
Jason d'Eon
University of Waterloo, 2019
PDF