Frequently asked questions

My research area is called Machine Learning, a field at the boundary of Computer Science and Statistics. It is very related to and largely overlaps with Data Mining and Artificial Intelligence. It involves designing efficient algorithms to make sense of data, discover its hidden structure. Machine Learning is also often described as the study of programs that improve with experience, i.e. as it gathers more data.

Here are some examples. Your email provider filters out spam messages automatically. How does the computer "know" that an article is spam? That's classification, a problem studied in Machine Learning and for which many algorithms (methods) were developed. Amazon recommends you a book based on the ones you've bought, again automatically. That's collaborative filtering. How long will you live, given how much you exercise, drink and smoke? That's regression. Based on many people's DNA, how many populations are there and how are they related? That's clustering. Find people in images, predict protein structure, detect suspicious credit card activity, translate speech to text, translate Spanish to English, fly airplanes automatically, explore Mars autonomously, etc. All these rely heavily on Machine Learning.

My work is fairly abstract most of the time. My work on Dirichlet processes had me thinking about probabilities on infinite objects for a long time. One day I told my wife I would try to classify news articles as a test application of my new algorithm. She looked at me in disbelief: "You mean you can use your research to do real stuff like that?".

Luckily, applications of Machine Learning such as spam detection and classification are becoming more well known so I can refer to them when I explain what I do. When I just say that my work is at the intersection of Mathematics, Statistics and Computer Science, I usually get "Yew, all the things I hate combined!", and when I say I study Artificial Intelligence I get "But isn't that all BS? Everybody knows computers are not intelligent."

No, I don't fix computers, nor would I know how to install your wireless. Computer Science is about manipulating information. What can you achieve by adding, multiplying, combining information, and moving it around? Once you've figured out a way to do something, say find the shortest path given two points on a map, you call your method an algorithm.

An algorithm is like a recipe. In this case the two points on a map are the ingredients and the shortest path is the finished product. Large algorithms are like giant recipes for entire banquets, and need to deal with many issues such as in what order you should bake the bread, meat, desert, how much storage you need for the utensils, etc. And if you have twice as many guests next time, will it take twice as long to prepare? I have two recipes, which one requires the fewest cooks? Which one is fastest? That's the kind of questions Computer Science studies.