tag:blogger.com,1999:blog-6555947.post24764356043505088..comments2024-03-14T01:32:43.610-06:00Comments on The Geomblog: Thoughts on ICDM I: Negative Results (part B)Suresh Venkatasubramanianhttp://www.blogger.com/profile/15898357513326041822noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-6555947.post-51410548309357726572011-12-21T07:46:13.924-07:002011-12-21T07:46:13.924-07:00For a more high-level take on the "power law&...For a more high-level take on the "power law" issue, you might check out my editorial: <br />The Future of Power Law Research<br /><br />http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.im/1150477668<br /><br />or my survey -- which (unfortunately) predates the SKG stuff, but a lot of the same issues keep reappearing.<br /><br />http://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.im/1089229510Michael Mitzenmacherhttps://www.blogger.com/profile/02161161032642563814noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-13591889942823098412011-12-20T17:46:33.198-07:002011-12-20T17:46:33.198-07:00Thank you very much! The Shalizi paper looks excel...Thank you very much! The Shalizi paper looks excellent.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6555947.post-83746721373914240672011-12-20T16:45:28.549-07:002011-12-20T16:45:28.549-07:00You should start here for some understanding of th...You should start here for some understanding of the controversy: http://cs.unm.edu/~aaron/blog/archives/2005/10/links_links_lin.htm<br /><br />In brief, the issue is "What's a power law", and I think the definitive research article on how to determine whether data follows a power law is by Clauset, Shalizi and Newman: http://arxiv.org/abs/0706.1062Suresh Venkatasubramanianhttps://www.blogger.com/profile/15898357513326041822noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-2915232971970292302011-12-20T16:34:12.005-07:002011-12-20T16:34:12.005-07:00It's interesting that you mention that there&#...It's interesting that you mention that there's some controversy regarding that. I'm only an enthusiast hoping to learn more. But John Hopcroft had a similar answer too. <br /><br />Thanks for pointing out that paper, I didn't click through. Looks very nice!Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6555947.post-27853150502931817842011-12-20T16:04:34.565-07:002011-12-20T16:04:34.565-07:00There's a fair bit of controversy about what e...There's a fair bit of controversy about what exactly the Barabasi/Albert graphs tend to produce in terms of degree distribution. I'm not knowledgeable enough to address that aspect of things: however, I point you to the Leskovec et al paper that I linked to above - they have a good discussion of desiderata for graphs that model social networks and how preferential attachment systems like Barabasi/Albert graphs fail to capture observed phenomena in social network graphs.Suresh Venkatasubramanianhttps://www.blogger.com/profile/15898357513326041822noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-63702872683852054262011-12-20T16:00:24.789-07:002011-12-20T16:00:24.789-07:00Nice post!
"the classic Erdos-Renyi (ER) mod...Nice post! <br />"the classic Erdos-Renyi (ER) model of a random graph doesn't have the heavy-tailed degree distribution that's common in social networks."<br /><br />What is your opinion of the Barabási-Albert random graphs in this context?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6555947.post-18127240229416271362011-12-20T11:02:44.521-07:002011-12-20T11:02:44.521-07:00From a more practical perspective, we also found t...From a more practical perspective, we also found that SKGs are incredibly sensitive to parameter tuning, and computationally very expensive to boot.Benhttps://www.blogger.com/profile/14936659851695940856noreply@blogger.com