[Computational Complexity] Innovation
- The new Innovations in Computer Science conference announced their accepted papers earlier this week including my paper with Rahul Santhanam "Bounding Rationality by Discounting Time". Shiva is collecting PDF pointers (hope to get ours posted and on that list soon).According to Noam, "this is the most interesting list of accepted papers that I’ve seen in years". Suresh seems happy too but some of his commenter's were less impressed. I view ICS as playing an orthogonal to STOC/FOCS, trying to present papers with potentially interesting ideas that wouldn't normally be accepted into STOC or FOCS.Why does STOC and FOCS not accept many innovative papers, even after adding the line "Papers that broaden the reach of theory, or raise important problems that can benefit from theoretical investigation and analysis, are encouraged" to the CFP? Since I can remember, STOC has focused more on solving well-defined problems with a deference towards technical proofs. A more innovative papers, that is a new model with a few simple results, tends to be harder to sell as one needs strong motivation and a model that really captures that motivation. When we do have more innovative papers they tend to be selected based on the authors more than the content.
But this problem has gotten worse in recent years as our field has grown and PCs seem unwilling to trade a solid technical paper for a risky innovative one. When the PC meets, an innovative paper has a higher standard, the need to justify a new model or problem. Papers that answer old open questions or extend earlier work can justify their models simply because those models have appeared in earlier conferences. But no theoretical model can exactly capture a real world scenario so one can always find negative things to say about any innovative model, causing more traditional papers to often win the fight for those last few acceptance slots.
ICS did not accept as wide a spectrum of papers as I would have liked, probably due more to lack of a broad submission base. And the models of most of these papers will likely go nowhere but the hope is some will go to instigate new research areas.Given the accepted papers can we tell yet if ICS is a success or a failure? Not yet and that's the point. Wait five years and see what fraction of papers from ICS (compared to STOC/FOCS) are still relevant and then we'll know.
Posted By Lance to Computational Complexity at 11/05/2009 06:21:00 AM