Annotated POPL 2007 Program

POPL 2007 is just around the corner, and the Little Calculist has taken time off from inventing Web 4.0 to track down all the papers available online, and collect them in an annotated program, Favourite titles: Proving That Programs Eventually Do Something Good, and Lazy Multivariate Higher-Order Forward-Mode AD. I have no idea what the later means, but I love the pile-up of jargon.

By Noel at 2007-01-15 09:55 | Theory | other blogs | 7257 reads

Automatic Differentiation

AD = Automatic Differentiation. It's a readable paper about an implementation strategy for multivariate AD that doesn't have "differential capture" problems. sigfpe posted in his blog a while ago really enlightening accounts of the technique, starting here, following up to differential geometry and ending by relating it to derivation of containers. I can't recommend sigfpe's blog highly enough.

By Matias Giovannini at Tue, 2007-01-16 02:58 | login or register to post comments

Related to the Non-Standard Interpretation paper

I spoke with the authors of this paper, and apparently their AD system is based on a technique that was the topic of a second POPL2007 paper: non-standard interpretations by opening closures.

I found the non-standard interpretation paper more interesting, frankly, since it seems to be a general approach that could be applied to lots of interpretations. For example, functional reactive programming, quantum computation, and even some static analysis techniques. I've worked with a language that implements functional reactive programming in Scheme (called FrTime), and many of the techniques in the non-standard interpretation paper are very similar to the techniques used in FrTime. Having a general framework that supports a broad class of such problems is very interesting.

By Kimberley Burchett at Fri, 2007-01-19 02:35 | login or register to post comments