Since I recently got "official word" that one grant is "in", I feel I can finally talk about my grant proposals from last year on the blog. I often feel there's not much discussion of grants or the grant process in our community, so perhaps this (and the comments from others) will give some insight to graduate students or young faculty.
Despite, or because of, the stimulus-based funding bump at NSF, one of my 3 submissions last year got funded. Interestingly, and unlike several other times in the past, I was not dismayed or upset at the rejections; I found the reviews quite reasonable, and believe that the rejected proposals are likely to be funded after a revision. (In the past, I've had revised proposals accepted on the 2nd -- or 3rd -- try.)
Staying positive, the proposal that did get funded was my individual (small) grant to Algorithmic Foundations. Perhaps not surprising anyone, it's all about hashing. I was shocked to find that the proposal seemed to have been reviewed by SIX people. I don't think I've ever seen a proposal with so many reviews. (I thought 3-4 was the norm.) I think the proposal also had the best overall scores I've received on a proposal.
That's not to say the reviews were universally positive. The pluses the reviews gave were the connections being made between theory and practice, the appealing questions, and my history of work in this area. The surprise there for me was that a few of the reviewers explicitly mentioned my "history" in the area; generally, I haven't found this explicitly taken into account in reviews before, although I certainly suspect it is often implicitly considered. The main negative in the reviews was the concern that the work would be too incremental. As one reviewer put it (paraphrasing slightly): "... I feel the proposal is somewhat incremental in its intellectual merit... The proposed research problems largely fall into the category of extending results in a mature field... I would have like to see something more out of the box..." The reviewer still gave the proposal a good score, so I can't complain that they didn't see the merits in tackling the many interesting open problems in the hashing area, or the importance of pushing new results in the area out into the real world. And I can certainly accept the criticism as a reasonable one -- indeed, my concern about this proposal going in was that it didn't have the "out of the box", high-risk flavor that seems in vogue at some areas of the NSF. Perhaps it's easier to accept the criticism since the proposal did get funded, but overall I thought the reviews showed understanding and were well thought out, balancing the pluses and minuses reasonably.
(OK, there was also some of the most-annoying but standard review comment: "He doesn't explain how he's going to solve the problems he's proposed," which I always find remarkable. If I knew, I'd already be writing or have written the paper... On the other hand, even in this case, the reviewer(s) did not use this as an excuse to disregard the proposal -- as they sometimes do -- so again, I can't really complain.)
Not surprisingly, the work that people mentioned as being most interesting/out-of-the-box was the paper on Why Simple Hash Functions Work, which utilized a not-worst-case model, showing that one would get near-uniform hashing with weak hash functions as long as there was "enough" entropy in the data stream. There should be more applications of these ideas, and thinking it through is clearly something I'll have to be getting back to.