[Prev][Next][Index][Thread]

Re: MI: why?

Marcel Weiher <marcel@cs.tu-berlin.de> wrote:
> "Richard MacDonald" <macdonaldrj@worldnet.att.net> writes:
>
> The GenVoca/Predator project(s) at U-Texas have some interesting
> work on this.  As you've shown here, collection classes are
> actually defined by orthogonal feature vectors, leading to
> a potential number of classes that is exponential in the 
> number of different features (the actual number is lower
> because some combinations don't make sense).
> 
> They call this the library scaling problem.  Anyway, the
> approach they use is a form of code generation, but not
> as ad-hoc as code-gen usually tends to be.
> 
>   http://www.cs.utexas.edu/users/schwartz/

In a functional setting, Simon Peyton-Jones paper "Bulk Types with
Class" is good reading, too:

http://www.research.microsoft.com/Users/simonpj/Papers/collections.ps.gz


Neel
Follow-Ups: References: