Slashdot Mirror

If they are not exclusive routes, feel free to suggest others.

I did.

First, I allowed for the possibility I was wrong. Second, "peer review" is a process - not a stamp of quality.

Yet you reject that possibility at every turn. Also, peer review is a stamp a quality -- it is a process designed to establish a threshold of quality through the input of the reviewers. Journals may do so well or poorly -- TACL is fairly selective.

My opinion is based on my experience and the information I have at hand.

Logical fallacy -- appeal to authority. Also, what experience, pray tell? I've looked at what you've disclosed about yourself -- 0 involvement in academic publishing.

That article is pretty light on actual information...anyway, here's a paper about getting machine learning to recognize opportunities for "that's what she said":

http://www.aclweb.org/antholog...

http://bit.ly/1dgDo7d . Come on slashdot editors, do the legwork and link the article directly!

Come on, martin, do the legwork and link it directly. This isn't twitter and most folks are wary of shortened links; trolls love hiding their goatse and tubgirl links. I only clicked it because your UID is relatively low and you hadn't (yet) been modded down.

You're a cunt for linking to a URL shortener instead of the article directly. You're as bad as the shithead editors. If you're worried about goatse or similar, don't use some shit URL shortener.

http://aclweb.org/anthology/D/D13/D13-1181.pdf

There. Easy. Why couldn't you have done that you high-horsed cunt?

Actually, those rates sound high to me. The last two conferences I attended (this and this) industrial members didn't have to pay more than $400 if they registered early. Snacks & coffee were always available, and a few meals and drinks (yes, I'm still a grad student these things are important to me :) And all the proceedings are free to access online.

Maybe some of the money goes towards scholarships, but I doubt that all of it does. Anyway as a student it works well for me, I don't want to pay towards other students' fellowships, I'm not going to buy the IEEE dental insurance or whatever, and I don't need a glossy magazine in my department mailbox. Maybe that stuff would be more important if I were in a sexier field? I dunno.

For me, reading the article immediately brought to mind the argument as to whether thought is a function of language, or whether language is a function of thought. I think that it's perhaps the latter, but that might only be true for abstract ideas (I don't know... I've never read any philosophy or studies on this, but I have pondered it in idle moments on occasion). Do thoughts rely on language at any point? Do the abstractions rely or draw upon language? And if so, are the thoughts of a non-English speaker "different", in some way, to the thoughts of an English speaker? (I'm just using English as an example -- don't read anything more into it than that). Perhaps egocentrism is something to think about as well. An example that comes to mind is the concept of time (see here, here here, and also the Aymaran language. I wonder how this "conversion" from thought/abstractions to language/description/communication really works.

I think you want the link for the paper, rather than the slides.

Hi again.

I wholeheartedly agree that computational neuroscience is an exciting field with lots of potential. Indeed, as you assumed in your previous post, "if you can build [i.e. model mathematically] it, then you know it"; and that certainly applies to the field.

Neural network modeling (the mathematical theory part of this issue) are a bottom-up approach, while experimental neuroscience as a top-down approach. The former tries to figure out the forest by examining possible trees and groups of possible trees, and the latter tries to figure out the tree by examining extant forests. I'm not sure where you are in your course of learning, but I would identify 3 broad areas that I think you could probably (surprisingly enough) treat sequentially with some success:

1) A broad and hopefully moderately deep understanding of neural network theory including topologies and learning techniques
Read lots, play lots with both *pen/paper* and computer programming application. "Parallel Distributed Processing" by McClelland and Rumelhart is a cornerstone and I think the the two volumes' software is available on their site. Yes, it's a bit of a misnomer because their ideas are not really what the field of computer science would consider either "parallel" or "distributed", and the physiological speculation is a bit dated in a fast-moving field, but the basics are there. You'll need tons of "tools" to work with, from simple sequential networks to more complex recurrent networks, pattern recognition techniques, and thus a healthy tolerance for the precise mathematical analysis of these problems. For someone versed in classical computer hardware, there's the occasional analogy with combinatoric circuits, sequential circuits, and the like (but not so much with Von Neumann's memory); just don't go overboard.

2) With the basics in hand, time to look at the behaviors people have observed, and also the models people are working on. This is broad enough that there is always someone coming up with previously unexplored arrangements of the basic theories, even if the topological arrangements of their networks are relatively similar. You seem to be somewhere in (1) and (2) at the same time, which is fine and a more interesting way of learning, but I suspect the stereotype of computational neuroscientists as "scruffy" comes from people running before they can walk, or running without bothering to tie their shoelaces. With regard to your specific question, I think it is asked on new legs, and you would be interested in this if you are comfortable with its underpinnings and style:
"Cascade Models of Synaptically Stored Memories" (Fusi/Drew/Abbott)
I asked google and this is an easy place to read it (not my site):
http://neurotheory.columbia.edu/~larry/FusiNeuron05.pdf
There are other relevant (or at least interesting) things in that directory.

3) Cognitive neuroscience. It's a bit more experimental/observational than the other, theoretical, areas. For my little ternary classification, I'm grouping the extremely wide field of computational linguistics in here instead of in (2) because it has a particularly direct association with the observed phenomenon of natural language. Behavioral and physiological study, especially of non-healthy or damaged/lesioned brains is of course not really the same thing as the modeling you're interested in (we are the same in that regard, incidentally).

Computational linguistics (and machine learning/AI in general) requires such breadth and depth of knowledge that it's (IMO) the most interesting aspect of cognition after logic itself; it encompasses logic, and other neat things. There is TONS of relevant reading here, though it's only sort of browsable:
http://www.aclweb.org/anthology-new/
You'll find one paper that interests you, exhaust every leaf in its citation tree (web), and have enough ideas to jump to any other tree. It's enough for serious academic study, and way more than you'll need to get your feet wet.

Good luck and happy reading, programming, and whatever else you decide.

Just about every every college or university with a decent Computer Science program has people studying NLP (Natural Language Processing). Government agencies are probably the biggest source of grants for research, so DARPA funding this is nothing new. Additionally, NLP is just a sub-field in AI. AI has somewhat turned into a bunch of sub-fields that all relate to computers doing something "intelligent". Other areas of AI include computer vision, expert system development, machine learning...etc. There's a more "open" version of something like CYC(an Ontology) called WordNethttp://wordnet.princeton.edu/, lead by George Miller of Princeton's Psychology Department. You may be familiar with it. It is like a dictionary, but the important part isn't the definitions, it is the subconcept/superconcept(hyponym/hypernym) relationships among senses of words.

Applications for NLP are all over the place. Search engines, for example, use a limited amount. There is a professor at UCF http://www.cs.ucf.edu/ who has developed a system to look up answers to questions in an encyclopedia and respond (in sentences). It also crosses over with data mining, and uses machine learning very often. Here is a link to one of the biggest annual conferences on NLP: http://www.aclweb.org/

If this kind of research interest you, and you're a student looking for an area of study, Computional Lingustics is an (IMHO) amazingly rich field of study, sausage notwithstanding.

void CShameless:Plug()
{
If you're running OS X, check out theConcept for an example of statistical language processing in action.
}