<p>haha, it's okay! =) you answered my question anyway</p>
<p>My school paid for the entire trip to the conference (yay!), (what is the cylsf?) though some people had to pay themselves.
BayesGuy, would I be correct that you happen to have a liking for Bayesian analysis stuff? (I'm actually interested in looking into that stuff some time)</p>
<p>As for type theory, there is far more to it that simply rejecting type mismatches on primitives(the simply typed lambda calculus is the least interesting of all lambda calculi). For example, polymorphic functions/datastructures for free instead of creating specialized overloaded versions! (and no, I'm talking about something far better that C++ templates, especially since they pale beside Scheme or even Lisp macros in general for metaprogramming).</p>
<p>But I'd best talk about this some other time, in a more appropriate forum for such cool topics. (My research paper for this year actually touched on the untyped lambda calculus, as I described an algorithm to allow recursive functions be written in a language that has support for functions as first class data but no support for recursive definitions themselves). COOL STUFF!</p>
<p>Schemer, are you an alumnus or junior? You sound like a smart kid to me. If you're applying to RSI, you got a good shot considering it sounds like you know what you're talking about.</p>
<p>...unlike me :(</p>
<p>I'm a junior and an oddball at that(even at my school, Bronx Science).
Thanks for your positive thoughts as my outlook, Zogoto. </p>
<p>good luck yourself Zogoto!</p>
<p>I'd be ecstatic if I got in, both for the fact that it will probably be a great social experience (if the people at the ncsssmst conference are anything similar ) and I'll have the opportunity to possibly work with researchers in my exact area of interest </p>
<p>there are only at most 6-8 professors who are doing stuff even vaguely related to my interests in the area in which I live (New York), compared with tons in those particular topics in cambridge/boston, and working directly under someone seems less strenuous that going along independently and emailing various professors all over the country whenever a question arises.</p>
<p>oh, and I read the research literature for fun! (bring on the phd theses!)</p>
<p>anyone can sound intelligent if they start spewing esoteric vocabulary out of their mouth</p>
<p><em>high fives static</em></p>
<p>now that's just hostile.</p>
<p>I was assuming that BayesGuy had some general familiarity based upon his type mismatch joke.</p>
<p>I'll explain what I meant, it that was that issue.</p>
<p>the lambda calculus: a really elegant model for programming languages and hench computation that simply makes use of the notion of a function</p>
<p>the simply typed lambda calculus: the lambda calculus restricted so that functions can only take one type, this is like C's type system without casting (converting integers to floats, etc) and the void type</p>
<p>polymorphic functions/datastructures for free instead of creating specialized overloaded versions! </p>
<p>By this I mean there are more sophisticated versions that work generically, and their type contains a variable(that becomes a concrete type upon being used), ie a type mapping a type X to a type X could be the identity function. (f z = z)</p>
<p>(and no, I'm talking about something far better that C++ templates, especially since they pale beside Scheme or even Lisp macros in general for metaprogramming).</p>
<p>metaprogramming is basically automatic code generation/specialization. Ie if you noticed a repetative pattern in the code one was writing, then one could write a macro that is essentially a new bit of syntax that takes as its arguements that bits that vary between the repetative portions of code.(its like extending the compiler or interpreter without having to modify the source of the compiler or interpreter)</p>
<p>Is this about improving the typed nature of programming languages and improving memory efficiency?</p>
<p>Schemer: I used to be impressed by that kind of stuff, but not anymore. The only thing that impresses me is something tangible. Do you have any code I can use to see what you are talking about? </p>
<p>It seems to me that what you are talking about is an all encompassing type. The reason I use int and float, etc, is because I need extreme performance. If we use this lambda calculus type for real, for everything, than if I am dealing with bits and bytes, I won't get performance.</p>
<p>I think you should look into compiler/interpretor coding. It is not easy, and much of what you have said has already been addressed in detail.</p>
<p>Although your research is good stuff, how will it affect the real world? Many researchers are abandoning LISP, because some of the features that were so appealing are now available in other languages. Many programmers are not willing to leave the C-style language format.</p>
<p>I could easily start spewing of information on recurrent neural networks, and their possibility to employ long term memory backed by real time networks. But until I have solid examples I won't. At least show us some websites where we can read you paper and code, because I am tired of this crap from every CS guy talking big, but coding small.</p>
<p>schemer-as interesting as your project sounds, I don't understand it. </p>
<p>what. is. ncmssst?</p>
<p>Intelliot,yes! Exactly! Variants of the typed lambda calculus/ and in some cases the untyped lambda calculus, are used as the intermediate language of fairly cutting edge optimizing compilers.</p>
<p>Sagar, if you want to I would be willing to email you a copy of the paper, which provides a sample implementation for PLT Scheme. (I also wrote a toy implementation of an interpreter in Haskell that did the same thing, but I realized that it would be easier to just to make use of a preexisting implementation so I wouldn't have to kill myself to verify its correctness). </p>
<p>Some good places to start to read stuff in this area include
<a href="http://readscheme.org/%5B/url%5D">http://readscheme.org/</a>
citeseer.org
the church project at Boston University's CS department
The web pages of Greg Morrisett (not quite correct spelling) and Norman Ramsey
at Harvard
Zhong Shao and several others at Yale
Robert Harper among others at Carnegie Mellon (Robert Harper coauthored the Standard ML definition)
Simon Peyton Jones of Microsoft Research (One of the main people behind Haskell)
the sites of the PLT projects at North Eastern, Brown and Rice
Andrew Appel of Princeton (he has a very cool book he wrote in 92 called Compiling with Continuations, I really should reread it when I get the chance) , he's the person behind the SML-NJ implementation of Standard ML.
Benjamin Pierce at UPenn (he wrote a really cool, wonderfully written book called Types and Programming Languages)</p>
<p>thats just a few webpages that I recall at this moment in time.</p>
<p>Back to what you were saying Sagar, that project I completed earlier this fall serves the purpose of simplifying the implementation of a language with Higher Order functions (functions as first class data).</p>
<p>the usage of the type theory stuff is pretty much as Intelliot said, with the additional benefit of verifying that a program is error free with regard to invariants expressible by the type system. Some cutting edge research is actually to have type systems that can do so for low level memory manipulation (pointers and such). I believe Greg Morisett at Harvard and HongWei Xi at Boston University are among the people who have projects touching on that stuff.</p>
<p>Xinerz, the project i did was I devised an algorithm that can convert recursive functions into a equivalent form that isn't self referential.</p>
<p>take the example of factorial
factorial (x) = if (x = 0) then 1 else x * factorial (x-1)</p>
<p>if we can treat functions as values (we can pass them to other function, we can have the following)</p>
<p>F(factorial, x) = if (x= 0) then 1 else x* factorial(factorial, x-1)
and have factorial be defined as
factorial (x) = F(F,x)</p>
<p>you can find stuff related to this by googling for the Y combinator</p>
<p>schemer, do you got to columbia SHP? i think i might know you.</p>
<p>Yup, complex variable calculus?</p>
<p>Conner? or something like that..</p>
<p>schemer- yeah! i kinda get it!!! ... kinda. =\ that's okay, if i get in (hah, fat chance) you can explain it to me. if not, i guess i'll just be ignorant for the rest of my life on converting functions. =)</p>
<p>and you guys can call me xine.</p>
<p>Xine, there are TONS of resources online on the stuff if you ever care to look.</p>
<p>huagm, you got the first letter right! Were you one of the people who actually talked in the class?</p>
<p>schemer, if anyone here gets it, it's probably me. I'm somewhat of a CS nerd. I have to admit, I haven't heard of some of the terms you're throwing out. But I think I understand the idea behind it.</p>
<p>I hope to do some related research someday, but my high school doesn't offer such resources. I've learned a lot by teaching myself. I think I would like RSI, but I probably don't have the credentials to get into such a selective program.</p>
<p>Hi- will a 730 Math IIC work against me for RSI apps if I have solid research and good SAT I scores? If so, I'm considering doing standby on Jan 22 to retake. If I do this, can I just e-mail my scores to RSI if they improve and will they consider the new scores? Also, if I do this, should I indicate that I am retaking on the app so they know to expect new scores? Thx</p>
<p>I'm completely self taught too, its just I'm so addicted to the topics I'm interested in that I've been reading everything I been able to find online for free or on a few occasions order a book from amazon for well over a year now.</p>
<p>My school's resources didn't and won't come into play for my research.</p>
<p>If you want to get into that topic, try checking out those resources I mentioned and/or teach yourself a dialect of ML (a very good language for implementing this stuff)</p>