These are projects that took significant time, mostly work projects. See in-page links below.
Object recognition using capsules 2010
Automatic circuit routing flow 2010
Judging ranking quality by user click-through data Summer 2009
Move/Copy for Windows Live SkyDrive Summer 2008
Robotic closet light inspector Spring 2008
Making graphene Summer 2007
These are projects that are done on my spare time or for a course. This list is very partial. I picked those that are still on my computer .
Projects that I did previously for work, school or fun.
I have been working on object recognition using capsules for my thesis. This work has generated considerate interests in the group. People are working on applying the idea to speech recognition and energy models.
During my PEY at Granite SemiCom, I designed and implemented a schematic based automatic placement and route flow.
During my 2009 summer internship at Microsoft. I implemented a system on SharePoint server based on research by P. Radlinski. The point is to provide an alternative measure in addition to NDCG, precision etc. After I finished the implementation, I also carried out some experiments. Unfortunately, design documents are confidential. But here are the slides from my non-technical presentation given to around 50 people at Microsoft.
During my 2008 summer internship at Microsoft. I was on the (3 interns) feature crew for the move/copy feature. This very important and basic feature has been live since Nov 2008 and is used by millions of people.
In teams of 3, we built robotic closet light inspector which performed the best out of 20 other teams of 3. Here is our report.
I was charged with producing graphene samples when I worked at the van Driel lab. My contributions are acknowledged in this paper. I believe (but cannot prove) that I was the second person to do this in Canada after a person in UBC and shortly after the discovery of this method.
I trained a neural net that performed the best in my class. This is a 5 cateogry (cars, horses, ships, planes, trucks) based on the CIFAR dataset.
For the very enjoyable graphics course taught by Prof. Kyros Kutulakos (fall 2008). I did a ray-tracer.
For another very enjoyable first course in visual computing (winter 2009). I implemented a bunch of image processing algorithm.
Pipe line a CPU with 8 bits instructions in Verilog on Altera Cyclone. Fun project.
I wrote this back in Gr. 11 after I read about the Mandelbrot set. The code is here. Even with the verboseness of Java, it is surprisingly short. I later wrote a more sophisticated version using C# adding in colour and all that. This was the first time when I got really excited about mathematics.
Intead of doing classic backtracking, this solver uses several elimination heuristics and was able to solve most of the puzzles published on some Sudoku website. Again, I also had a C# rewrite that worked better, but cannot find it now.