About Myself

Artificial Intelligence? Computer Vision? Natural language Processing?

Artificial Intelligence never fails to appeal to me.

Just like Doraemon to Nobita Nobi , I have been dreaming of having a computer friend since I was a kid. It has a bunch of fancy inventions from the 22^nd century in his four dimension pocket, joins me on the adventure exploring this world, and most importantly, teaches by honest words and actions instead of banalities. For years, I scratched numerous drafts about how it would look like, imagined the fancy capabilities it might have, and even came up with scenarios where it could come to realities without scaring my parents. Well, most of them is lost in time, but the fantasy for building such a powerful program still sticks to my mind.

Finished the elementary school and high school in China, I was fortunate to came aboard to Canada to continue my study at McGill University, majoring in eletrical engineering. Studying at McGill really broadened my horizons, and it was the real starting point for me to get to know programming algorithms, electrical circuits, information theory and a lot of cool stuff. Being exposure to this massive amount of new knowledge is an exciting yet challenging journey, and no need for mentioning how many nights spent in the Schulich library. However, there was always one question confusing me: What do I want to do with these great pieces of knowledge? Trying to find this answer, I took a few internships across different industries, with various technical focuses. I witnessed how teams across the world leveraged their own experties, contributing to making a great product that faciliated millions of people. I was also amazed to see that people were willing to spend days in improving browsers just to render contents with 1 second faster. On one side, all of these experiences pushed me to go further with my study, and opened completely new worlds. On the other side, they also deepened my question. Time is limited, decisions need to be made.

In the last year of my bachelor study, one of my friends mentioned me a course named computer vision that he was about to take. For curiosity, I clicked the course website in the evening, and was greatly surprised of the amazing things that it can do. For the first time, I realized computer algorithms are not only applied to sort an array or to design the shortest path for a busy postman. Instead, it can be so connected to the most of vision-related tasks we perform in the daily life, to recognize a poker card, to restore an old image, and even to stabilize poor-taken videos from amateurs. I took the course without hesitation. It was also around the same time when everyone started to talk deep learning. The success of AlexNet proved that applying the data-hungry iterative process is feasible even in the high dimensional data like images. Millions of dollars were spent by the those tech giants to build AI labs. The power of supervised learning can even be explained clearly by hosts in late night TV shows。 It looked like everything could be solved by deep learning, and the only problem remained as how we can apply this fashionable approach to actually do them. Being fascinated to know all of this, I thought it would be cool to go one step further by taking the graduate school, in order to have a better understanding of how things exactly work. With a mixed feeling for being excited and concerned, I chose the graduate school at McGill over a job offer at a bank. Many years later when I looked back at this decision in those sleepless nights, I really appreciated the choice I made: It is always easier to believe a plausible statement from others, rather than figuring out if the statement is really valid by myself. The Deep learning, which sounds promising and does deliver astonishing performances, also comes with a lot of non-neglectable problems. Through my graduate study, one of my main tasks was to identify these limitations and designed corresponding solutions to mitigate them.