If you are not already familiar with deep generative models, we’ll be working on generative models covered by lectures in this playlist.
Application Requirements for OoDD
Sorry for higher than usual requirements for this project, however as research projects move fast (we are competing with the entire field) and involves substantial commitment from everyone, this is necessary and this is being responsible for everyone on the team.
Background : Minimum GPA of 3.5/4.0 (3.7+ is preferred). Familiar with pytorch. Good coding style and code review experience. Knowledge in deep generative models is a plus. Software engineering experience is a double plus.
Instructions: Email your resume/CV and transcript to Sheldon to huang at cs dot toronto dot edu or contact Sheldon on Discord. You can expect a coding chanllenge and an interview for the application process.
Expected commitment: It’s okay to not know much about this specific topic but this is a serious research project so we’d expect at least 10 hours per week commitment on this project to learn and to code. You can expect to learn a lot by asking a lot of questions and coding a lot. What could be a better way to learn?
Application Due: Saturday, September 19, 2020
How it works
This project will be led by Sheldon and potentially some of his PhD supervisors, and just like all research projects, there will be publication opporunity, your name will be on the authorlist depending on your level of commitment.
You will also potencially collaborate with researchers at FOR.ai. We will have a biweekly meeting. We will communicate on a private group chat on discord and collaborate on a private github repo using the codebase detailed at the bottom of this page.
Road map
You can think of this project to be based on 5 paper reproducing project. Each team member will take charge of reproducing one paper while also helping each other with code review (so that bugs will be catched later on. Everybody writes bugs.) And members will also be working on some shared components together. Once we have a paper reproduced, we can clean up the code and separate out only this paper for a public release. The research project is based on those methods. We will aim for a submission once we have significant results.
