tag:blogger.com,1999:blog-8890204.post91439090567585902..comments2024-03-10T05:26:42.148-04:00Comments on My Biased Coin: The Old SODA vs. FOCS/STOC DebateMichael Mitzenmacherhttp://www.blogger.com/profile/06738274256402616703noreply@blogger.comBlogger17125tag:blogger.com,1999:blog-8890204.post-9918863122909885902007-07-19T17:36:00.000-04:002007-07-19T17:36:00.000-04:00According to web site above WADS at 171 is better ...According to web site above WADS at 171 is better than SoCG at 175. So much for citation counts.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-44015559217759136382007-07-19T16:11:00.000-04:002007-07-19T16:11:00.000-04:00Well according to http://libra.msra.cn/default.asp...Well according to http://libra.msra.cn/default.aspx the best CS conference is SOSP (STOC is number 10). . <BR/><BR/>The best CS journal is TOCS (Transactions on Computer Systems, JACM is no. 7 after CACM). <BR/><BR/>The best author is Ullman (Karp is #15, Wigderson #187).Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-77345647208051835712007-07-19T13:10:00.000-04:002007-07-19T13:10:00.000-04:00ARV == Arora, Rao, Vazirani. O(sqrt(log n)) approx...<I>ARV == Arora, Rao, Vazirani. O(sqrt(log n)) approx for sparsest cut (and best paper in STOC'04).</I><BR/><BR/>The paper still doesn't ring a bell, though the accomplides are well known. <BR/><BR/>Widely used theory acronyms based on initials: RSA, L-Z, AVL, BMP, any others?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-76192469452061925232007-07-19T11:32:00.000-04:002007-07-19T11:32:00.000-04:00ARV == Arora, Rao, Vazirani. O(sqrt(log n)) approx...ARV == Arora, Rao, Vazirani. O(sqrt(log n)) approx for sparsest cut (and best paper in STOC'04).pálenicahttps://www.blogger.com/profile/15057290951441022430noreply@blogger.comtag:blogger.com,1999:blog-8890204.post-91335345848678947852007-07-19T09:03:00.000-04:002007-07-19T09:03:00.000-04:00To "theory grad student": I can guess what DDN is ...To "theory grad student": I can guess what DDN is (though I doubt anyone outside of crypto will be able to --- and I wonder how important this paper is outside of the crypto community, but that is a question for another day), but what is ARV?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-82919759971865251112007-07-18T22:36:00.000-04:002007-07-18T22:36:00.000-04:00There're a couple qualitative issues to bear in mi...There're a couple qualitative issues to bear in mind:<BR/><BR/>1. "vision"/"ground-breaking" papers are likely to be submitted and accepted to focs/stoc. these include papers that introduce areas, definitions, fundamental techniques. [think ARV, DDN, etc.]<BR/><BR/>2. a solution to an open problem posed in an IPCO paper is unlikely to especially interesting. by that reasoning, we'd want to relate our work to other focs/stoc papers in a soda/focs/stoc submission.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-51650116682598908312007-07-18T22:35:00.000-04:002007-07-18T22:35:00.000-04:00What possible lessons can we learn using these cit...<I>What possible lessons can we learn using these citation tools?</I><BR/><BR/>Given anonymous Jul 18 12:57 little experiment, probably the answer is "not much". I doubt that they give much information even in the aggregate. Anyway, before starting playing with the numbers one would have to try to establish this. <BR/><BR/>This seems to me a very bad way to evaluate conferences. Probably the single best strategy to increase citation counts would be to publish biology papers.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-33888389251448332802007-07-18T20:08:00.000-04:002007-07-18T20:08:00.000-04:00To build on the comment by anonymous (July 18, 200...To build on the comment by anonymous (July 18, 2007 12:26 PM), there are at least two potential endogeneity problems here -- in addition to the increased visibility of STOC/FOCS after the fact, there is the effect that researchers might want to send papers first to the venues where they feel they will be more visible...maybe someone wants to check on the fates of SODA papers that were first rejected by STOC/FOCS? ;) (Not that that would solve the first problem...)Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-50955695123073948292007-07-18T13:00:00.000-04:002007-07-18T13:00:00.000-04:00More here, as well as some raw data to play with.More <A HREF="http://geomblog.blogspot.com/2007/07/soda-vs-stocfocs.html" REL="nofollow">here</A>, as well as some raw data to play with.Suresh Venkatasubramanianhttps://www.blogger.com/profile/15898357513326041822noreply@blogger.comtag:blogger.com,1999:blog-8890204.post-37087509947091481752007-07-18T12:57:00.000-04:002007-07-18T12:57:00.000-04:00Re: quality of citation count.Here's a data point:...Re: quality of citation count.<BR/><BR/>Here's a data point: my top eight papers quality-wise --according to my own very subjective opinion-- are number 11, 2, 9, 1, 14, 8, 12, and 13 (in that order) by citation count. IMHO numbers 3, 4, 5, 6, 7 and 10 by citation count are much inferior work, but they just happen to be early work in areas that are currently hot. <BR/><BR/>Observe that the average rank by citation count is 8.75 while the average rank by quality count is 4.5. The noise factor is then (8.75/4.5-1)*100= <B>94%</B> which is close to the 80% difference between STOC and FOCS noted by the previous post.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-70285477355914074342007-07-18T12:48:00.000-04:002007-07-18T12:48:00.000-04:00Citation counts are very rough proxies for qualit...Citation counts are very rough proxies for quality. I wonder if the level of resolution is good enough to distinguish between 14 (SODA) and 21 average citations (STOC). In fact the disparity between STOC and FOCS which are rather similar in scope and quality suggests that the noise factor of citation counts is upwards of 80%.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-57441869341621753552007-07-18T12:26:00.000-04:002007-07-18T12:26:00.000-04:00Could it be that there is some reverse causality h...Could it be that there is some reverse causality here? It could be that FOCS/STOC papers simply have more visibility, being more well-regarded conferences. If this caused some valuable papers to go unread by as broad an audience, whereas FOCS/STOC papers get built upon more often, you'd get a lower citation count for SODA.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-34376734116192099692007-07-18T11:32:00.000-04:002007-07-18T11:32:00.000-04:00It is not surprising that a conference accepting 6...It is not surprising that a conference accepting 66 papers across a broader portion of the TCS spectrum dominates an algorithms oriented conference with 122 acceptances. The real question is if there is anything gained from the extra selectivity of STOC and FOCS. <BR/><BR/>Say we could create a conference that would publish only the single best paper that year (like Aaronson's fictional <A HREF="http://scottaaronson.com/blog/?p=253" REL="nofollow">FOCS'36</A>). Such a conference would blow away your median cites and max cites count, but as a <B>conference</B> it would be awful: few people would attend, one would get a rather limited perspective of all recent relevant work, there would be little opportunity for followup work, with low probability of the single paper being relevant to your work, etc.<BR/><BR/>Then it seems to me that what most people implictly contend when they include SODA in the big three is that as a conference SODA is equally if not substantially more successful than the modern incarnation of STOC/FOCS. <BR/><BR/>To wit, SODA seems to attract most of the researchers in algorithms (including stars and substantial numbers from overseas), it has broad coverage (as opposed to overly narrow) of the intended areas, almost any result worth noting in algorithms will appear there and as such, regardless of one's research area, there always are several sessions of one's own interesting with papers well worth listening to, among others, and overall the quality of the papers is not that much below of STOC/FOCS.<BR/><BR/>This is why SODA has claims to being part of the big three. For the good of the field I hope it never goes the ultraselective route of STOC/FOCS which has turned them into nearly conferences which only authors attend.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-71124402741282634822007-07-18T09:09:00.000-04:002007-07-18T09:09:00.000-04:00dang, at first i thought you were using "dominated...dang, at first i thought you were using "dominated" as a verb, not an adjective.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-85081843041384876312007-07-18T08:51:00.000-04:002007-07-18T08:51:00.000-04:00I'd look at scott's blog, the entry about the Turi...I'd look at scott's blog, the entry about the Turing article.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-35808112452152684922007-07-18T03:11:00.000-04:002007-07-18T03:11:00.000-04:00It is interesting to try to evaluate each conferen...It is interesting to try to evaluate each conference as a whole, and it has some limited value (for a department chair deciding whether to fund travel, or for a newbie trying to decide where to send their work, or where to read to learn about a field), but there is a dismaying tendency to evaluate a piece of research, or a researcher, by using the conference where it appeared as a proxy. Once work is say 5 years old, it should surely be evaluated by the number and type of citations to that paper, or similar paper-specific measures, rather than by measures of the conference as a whole.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-8890204.post-54449071789036303812007-07-17T19:55:00.000-04:002007-07-17T19:55:00.000-04:00Comparing FOCS/STOC to SODA is a bit like comparin...Comparing FOCS/STOC to SODA is a bit like comparing apples to oranges, and it could very well be the case that <EM>both</EM> FOCS/STOC are better theory conferences than SODA, while SODA is a better algorithmics conference than FOCS/STOC.<BR/><BR/>The main difference with SODA, in my opinion, is that it is rather narrow in scope. As a cryptographer, I find lots of interesting papers in STOC/FOCS (even beyond the small handful of crypto papers accepted), while I rarely (if ever) have occasion to read a SODA paper. Just based on scope alone, I don't think you can call SODA one of the "big three" theory conferences, even if its quality may be on par with STOC/FOCS.<BR/><BR/>SODA also accepts way more papers than STOC/FOCS. Whether this is good or bad is irrelevant here, but it surely dilutes the average paper quality.<BR/><BR/>For a more meaningful comparison of the citation counts, you may want to compare citations of the <EM>algorithms</EM> papers in STOC/FOCS to the papers in SODA.Anonymousnoreply@blogger.com