# Monthly Archives: November 2012

## How do random forests work?

I’ve not had much interaction with the Q&A site Quora before, instead using stackoverflow or more recently biostars for bioinformatics questions, but while gathering sources for a plain english explanation of the machine learning technique called Random Forests, I can across this great and highly intuitive analogy posted by Edwin Chen. I’ve since decided I won’t be able to do better, so here is Chen’s answer to the question:

### How do random forests work in layman’s terms?

Suppose you’re very indecisive, so whenever you want to watch a movie, you ask your friend Willow if she thinks you’ll like it. In order to answer, Willow first needs to figure out what movies you like, so you give her a bunch of movies and tell her whether you liked each one or not (i.e., you give her a labeled training set). Then, when you ask her if she thinks you’ll like movie X or not, she plays a 20 questions-like game with IMDB, asking questions like “Is X a romantic movie?”, “Does Johnny Depp star in X?”, and so on. She asks more informative questions first (i.e., she maximizes the information gain of each question), and gives you a yes/no answer at the end.
Thus, Willow is a decision tree for your movie preferences.

But Willow is only human, so she doesn’t always generalize your preferences very well (i.e., she overfits). In order to get more accurate recommendations, you’d like to ask a bunch of your friends, and watch movie X if most of them say they think you’ll like it. That is, instead of asking only Willow, you want to ask Woody, Apple, and Cartman as well, and they vote on whether you’ll like a movie (i.e., you build an ensemble classifier, aka a forest in this case).

By using this ensemble, you hope that while each of your friends gives somewhat idiosyncratic recommendations (Willow thinks you like vampire movies more than you do, Woody thinks you like Pixar movies, and Cartman thinks you just hate everything), the errors get canceled out in the majority. Thus, your friends now form a bagged (bootstrap aggregated) forest of your movie preferences.

There’s still one problem with your data, however. While you loved both Titanic and Inception, it wasn’t because you like movies that star Leonardio DiCaprio. Maybe you liked both movies for other reasons. Thus, you don’t want your friends to all base their recommendations on whether Leo is in a movie or not. So when each friend asks IMDB a question, only a random subset of the possible questions is allowed (i.e., when you’re building a decision tree, at each node you use some randomness in selecting the attribute to split on, say by randomly selecting an attribute or by selecting an attribute from a random subset). This means your friends aren’t allowed to ask whether Leonardo DiCaprio is in the movie whenever they want. So whereas previously you injected randomness at the data level, by perturbing your movie preferences slightly, now you’re injecting randomness at the model level, by making your friends ask different questions at different times.

And so your friends now form a random forest.

So I have now been convinced that Quora is worth joining!

Filed under Machine learning

## BBC BSc

The BBC has impressed me recently with a couple of standout science-related programs that haven’t shied from some traditionally dry subject material.

Two weeks ago there was The Hidden Life of the Cell — a grandiose, narrative-driven tale of the epic struggle between the human immune system and viral invaders. It showcased some really impressive visualisations of molecular biology, as well as some interesting insight from scientists like Nick Lane* of UCL and narration by none other than Doctor Who himself!

More recently, Dara Ó Briain kicked off his new series Science Club. Dara is no stranger to hosting popular science shows, having helped out Brian Cox with Stargazing Live (didn’t see) and hosting the maths-based series  The School of Hard Sums (didn’t enjoy). Viewers of the latter program will recognise a similar format with Science Club, but this time the Irishman’s charm is lacquered upon more general topics, starting with reproduction and inheritance. It was nice to spot some different (well, lesser-known) faces and I appreciated the brave foray into epigenetics, which I expect many similar programs would avoid at 9pm on a Tuesday; nevertheless, it seemed to go down well with some:

Both of these shows are nice departures from those that typically fill the “Science & Nature” category of iPlayer — namely Countryfile and AutumnWatch — and I’d like to think they reflect a growing interest of the populous in learning real science. While I do have criticisms of both the shows I mention here, kudos to the BBC for putting out these programs (especially at primetime) and I look forward to more quality science programming.

* Good scientist, gaudy website!