KDnuggets : News : 2007 : n08 : item4 < PREVIOUS | NEXT >

Features

From: Gregory Piatetsky-Shapiro
Date: 23 Apr 2007
Subject: Interview with Simon Funk: What attracted you to machine learning?

GPS: Q1) What first attracted you to data mining / machine learning and what education did you receive?

Simon Funk
Simon Funk
Laziness! Seriously, laziness is behind many of my projects and machine learning is the ultimate expression of this. (Never mind how many years of hard work it may take to automate a ten minute task...)

I recall early on in my undergrad career trying to learn Lisp with a friend of mine and rapidly losing interest when we saw how much typing was involved. And not just the parentheses but everything between them too. I think about two hours in, I turned to him and said "I bet there's a tiny little program that would fit on a page and would learn all this stuff on its own."

This seemed as good an excuse to quit as any, so we went out for pizza and forgot about Lisp. Not long after that I coincidentally met a guy who happened to be doing his thesis on neural networks (something I knew nothing about) and we got to chatting and he was telling me about this one page of code that learns all this stuff on its own... And of course my brain is going "Aha! I knew it!" which no doubt created deep Pavlovian associations forever endearing to me neural networks and their ilk.

Within a couple of weeks I'd trained up a net to recognize musical instruments by sound, thus solving the dreaded question of what I was going to do as a project for my computer music class...

And so a great partnership was formed: I write a few lines of code; the computer does all the work. I eventually received a B.A. in computer science from UCSD, with minors in math and psychology. In retrospect this was a perfect trio for machine learning, though in truth I picked them because they maximized overlap and minimized blue-book exams...

Part of that overlap was a psychology class taught by V.S. Ramachandran on vision and optical illusions, which was a great intro to the idea of perception as a mechanism. I also finagled my way into Robert Hecht-Nielsen's graduate neural network seminar where I got quite a broad exposure to the field over the course of the year. (Though largely I just remember sitting back and listening to him tell stories, and marveling at his keen taste in pocket kerchiefs.) At the time, those guys didn't have the notoriety they do now, but they were both genuinely enthusiastic about the material, and it was interesting stuff, so those are the classes I remember. Who knows--maybe if my hydrodynamics prof had been as enthusiastic I would have become a plumber.

Bookmark using any bookmark manager! What's this?


KDnuggets : News : 2007 : n08 : item4 < PREVIOUS | NEXT >

Copyright © 2007 KDnuggets.   Subscribe to KDnuggets News!