It took a little while after the acceptances went out, but now the lists of accepted papers are up for both Sigmetrics and EC for this year’s conferences. Our group was lucky enough to get 3 Sigmetrics papers and 4 EC papers in this year, so Caltech will be strongly represented at both places once again!
From the looks of it, both conferences will be quite interesting, and I’m definitely looking forward to attending them in June… Happily, they are taking place during back-to-back weeks, rather than during the same week like last year. (Many of us bounced back and forth between them last year, which was only possible because Sigmetrics was in Pittsburgh and EC was in Philly.)
One thing I’ve talked about already in an earlier post is the growing overlap between Sigmetrics and EC among young researchers, e.g., nearly all of the recent Sigmetrics Rising Star award recipients have some ties to the EC community through the Network Economics field. Happily, this year, the accepted paper lists highlight that the connection, though still weak, is growing.
There are a number of topics where both conferences will have a significant focus:
- Auctions: Of course, EC is dominated by work on mechanism design. While this has not traditionally been true at Sigmetrics, there are a number of papers this year that are looking at auctions in the context of cloud and energy markets.
- Epidemics: Both conferences have a significant subset of work on epidemic models, and more broadly, social network characterization. However, from the titles and authors, my guess is that the toolsets being used in the two conferences are quite different.
- Pricing: Pricing is perhaps the most classical econ-related topic that shows up in Sigmetrics, and this year, there are papers looking at pricing models in both network economic settings and power systems (2 of which come from Caltech). And, of course, there are a lot of pricing papers at EC, too.
- Matching markets: Interestingly, though matching markets show up in both venues, the motivations are quite different. In EC, the story is usually related to kidney exchanges or ad auctions, while in Sigmetrics, it is usually related to cloud computing and scheduling. The models and results are quite related, though, so I hope the authors look at each others’ work!
So, there are lots of areas of overlap, but it is certainly true that the bulk of the two conferences are quite different.
EC is still overwhelmingly dominated by mechanism design. But, I’m happy to see that the program is a bit more varied this year than in the last few. For example, there is a much larger fraction of papers this year that is coming from the Network Economics community (which is a mix of CS / EE / Control folks) as opposed to the Algorithmic Game Theory community (which is mainly CS folks).
Sigmetrics, on the other hand, is still dominated by application-oriented questions. Interestingly, though, the applications driving the conference are changing rapidly now. Not long ago, networking dominated Sigmetrics, but these days, there is a varied mixture of applications, almost evenly divided among networking, energy, cloud, social nets, memory, and a smattering of a few other topics (including privacy for the first time). Also, while the theory side of Sigmetrics has traditionally been dominated by stochastic analysis of queueing, this is no longer the case. There is still some classical work on this topic in the program, but the bulk of the theoretical work is now involving optimization, control, and learning. In fact, the growth in the amount of work at the boundary of optimization and learning is probably the biggest trend visible in this year’s program.