Yahoo! have* posted their list of key scientific challenges in machine learning. I don't work on and hardly know anything at all about any of these topics. In fact, I think I understand what the question is in only three out of five cases.
Funny. I've always seen myself as working on some sort of machine learning, using computational intelligence methods. But if this is machine learning, I'm certainly not working on machine learning - it's about as related to my work as meteorology or linguistics is. So I should probably not say that I work on machine learning any more than I say that I work on meteorology or linguistics.
I'm actually OK with this, as I can still claim that I'm a computational intelligence researcher. Good enough for me.
But still... who gets to set the agenda? Ten years ago, what I do was machine learning; at least if Tom Mitchell's book is anything to go by. Nowadays, the important "machine learning" conferences such as NIPS and ICML wouldn't even look at the sort of stuff I do, irrespective of its quality. This is mildly annoying, as these conferences somehow have more prestige than CEC, Gecco and PPSN (probably because of ridiculously low acceptance rates).
And, most importantly: how does this semantic drift affect who gets the grant money?
* My intuition is really to write "Yahoo! has posted" here, as Yahoo! is a corporate entity usually referred to as it rather than they. However, British English seems to want to have it otherwise.
This comment has been removed by the author.
ReplyDeleteOne of the topics does specifically mention something that we have been talking about investigating, namely "How do you reuse knowledge bases & ontologies in machine learning?"
ReplyDeleteAnd in your blog post, I at first read mereology rather than meteorology.
Yahoo is a private entity. If they know / think that those things would bring them money than it makes perfect sense for them to want to work on those. Isn't that a normal attitude?
ReplyDeleteHi Julian,
ReplyDeleteI think Yahoo is referring specifically to *Statistical* Machine Learning (most of their problems involve classifiers of some description).
CI methods have certainly been used in this area before (NNs being the most effective AFAIK).
Perhaps you just specify the field your work is applied to in order to avoid confusion.
Marie: didn't think about that. We should talk about that again.
ReplyDeletegiures: Yes?
Daniel: You're right, and this is how I used to think of the matter. However, it seems that e.g. the NIPS and ICML committees, and Chris Bishop, only think that the statistical perspective is interesting. And they pretty much define the field. It is possible to get CI-ish neural networks papers into NIPS, but then you have to hide the fact that they are actually about neural networks (e.g. "differentiable function approximators" seems to be acceptable). So in practice, it seems that the statistical guys own the machine learning name nowadays.
Hi, Julian!!!
ReplyDeleteHehehehehe! Funny challenges those ones! :D
It's funny: When someone tries to set up a list of top something (facts, songs, challenges, whatever...) the bias towards the author's own area of interest seems almost inescapable.
There is an evolutionary computation book named "Frontiers of Evolutionary Computation" which lies on that category. After reading some pages and wander through the book, I got the insight that despite all those stuffs those authors were discussing, the most known and used evolutionary algorithm still is the good old elitist SGA! It dates back to the early 1970s (or late 1960s)!
So, the book's name should not be "Frontiers of Evolutionary Computation" but "Guess What??? We Are Still Using The Elistist SGA." At least it would be much more honest, I think. :)
So, the Yahoo post has nothing to do with "Key Scientific Challenges In Machine Learning", but "Key Information Technology Problems Yahoo Has Faced".
The EC book I aforementioned is just a sample of a, let's say, post-modern times' phenomenon: Edited books. Springer has lots of them, coming from the most obscure subjects and, not so rarely, about areas that I doubt there is a nice amount of interest on them. I hope well regarded journals do not follow that fad, otherwise soon we will see "IEEE Transactions On Telephone Cabin" and so on.