And thus ends the hell week. Three assignments handed in, and a presentation for my reading group. Helped with frog experiments, sent out the newsletter for DigitalEve Japan, and played softball. I thought it would all kill me, but I didn't even have to pull a tetsuya (all-nighter). Yatta!
I absolutely love the reading groups we do in our lab. We're reading Pattern Recognition and Machine Learning, also lovingly known as "PRML". Indeed, it's a tough book to get through, and "doing a presentation" is not just slapping together some slides about a topic you already know. To present a chapter from the book, you really have to:
Step 1: Read the chapter (how many times did I fall asleep...)
Step 2: Understand the chapter (deciphering an incomprehensible language has become my specialty)
Step 3: Decide what's important (less is more)
Step 4: Make the slides (pictures of bunnies help)
Also, for every single presentation I do, I start with a "hook". It's the most important part of a presentation. If you don't tell the people why they should listen to you, they won't.
Typically I start with a problem targeted to my audience. "Here's problem X. Has this ever happened to you?" They start nodding, and no matter how inane the solution, they will listen. Works every time.
Anyway, I'm just happy because, although my slides were in English, I managed to present in Japanese! It was the 2nd time I tried doing a talk in Japanese, though I had to fall back to English a few times. And, despite the language barrier, more than one labmate told me it was easy to understand. Yay!
Doesn't the bunny make EM for GMM look less scary? I think so.
Edit: Here is my full tutorial on Expectation Maximization and Gaussian Mixture Models. It also includes an overview of K-means. All pictures (except for the bunny) are from Christopher Bishop's Pattern Recognition and Machine Learning.