Sunday, May 15, 2011

NSF Workshop: 8F (Algorithms in the field)

I'm off to DIMACS for the NSF workshop 8F ("Algorithms in the field", AITF, get it?):
Computer Science (CS) is rapidly evolving. We need to understand the evolving CS systems and their algorithmic foundations. This needs close collaboration between the researchers in algorithmic foundations and expert researchers in various areas. The premise is that researchers from different communities should work jointly “in the field”, constantly informing each other, inventing in their respective areas, and forging systems that are simultaneously validated by both algorithmic and systems communities (and without needing to make independent research threads meet ex post).

The purpose of this workshop is to provide a working vision for examples of Algorithms in the Field, near term goals and directions for research in the future. The outcome of this workshop will be a report contributed by the participants that will inform the academic community and future funding programs. Scientifically, we hope bringing people with widely varied research interest together with algorithms researchers will lead to happy collisions, and unexpected directions of research.
There's a great mix of researchers from within theory and from more applied areas, and of course the topic is near and dear. Along with a long list of talks, the idea is to have a number of breakout sessions on different topics in this area. One that I'm involved with is 'Core Algorithms', whose mandate (roughly speaking) is to figure out how basic algorithms research might evolve over the next many years, in relation to events in the larger arena of computing research and technology.

Of course it's a mug's game to make predictions about the future of technology, or worse still, about the future of research. Unlike many other subfields of CS, theoryCS doesn't really lend itself to long-term prognostications or discussions of the 'this is what we should work on' variety.

But from a funding perspective, if we can identify interesting directions to explore even if we look completely stupid in retrospect, this could be a fun exercise. One framing device that we might try to use is:
View core algorithmic developments in two directions. In one direction, there are many concepts from the world of theoryCS that slowly make their way into the more applied realms. In this context, one might ask about which paradigms are either ripe for "tech tranfer", are likely to do so within a few years, or need more 'baking' before they're ready, but have potential.

    The reverse direction is the idea that core algorithmic questions and models are motivated by applications, whether they be models of computation like external memory/streaming/multicore/mapreduce, or conceptual developments like smoothed analysis, complexity classes for iterative algorithms, numerical algorithms and precise computations, and so on.

    In this context, interesting points for discussion would be: what kinds of paradigms do we see approaching from "applied land" and how might they influence the kinds of core algorithmic questions we ponder.
 If you're at the workshop, you can make your opinions known. But even if you're not, you can voice them right here ! I'll incorporate comments posted here into the discussions (and sign your name if you'd like credit). 
Post a Comment

Disqus for The Geomblog