tag:blogger.com,1999:blog-6555947.post8211560338624490290..comments2014-01-12T10:46:48.153-07:00Comments on The Geomblog: Clustering: The "I don't like you" viewSuresh Venkatasubramaniannoreply@blogger.comBlogger8125tag:blogger.com,1999:blog-6555947.post-32895751392937762662009-07-02T06:55:27.352-06:002009-07-02T06:55:27.352-06:00very interesting,
How do you compute the number o...very interesting,<br /><br />How do you compute the number of voronoi partitions (in the case centers are not from the dataset)? A reference would be appreciated.nicolaenniohttp://www.blogger.com/profile/11858065497262185094noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-60643674952034317302009-06-28T22:14:51.757-06:002009-06-28T22:14:51.757-06:00Thanks for this very useful series, Suresh, and I ...Thanks for this very useful series, Suresh, and I look forward to the rest of it. Meanwhile, could you or your readers point me to some publicly-available datasets on which we can run clustering approaches? Perhaps this info can be posted here, to benefit all readers. Thanks in advance.aravindhttp://www.cs.umd.edu/~srinnoreply@blogger.comtag:blogger.com,1999:blog-6555947.post-36708446555053888962009-06-28T11:43:36.193-06:002009-06-28T11:43:36.193-06:00yes, I'm assuming that the center might not be...yes, I'm assuming that the center might not be from the data set. Else, you are right in that the choice of k centers upper bounds the total number of partitions.Sureshhttp://www.blogger.com/profile/15898357513326041822noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-21221452353677754742009-06-28T11:23:38.631-06:002009-06-28T11:23:38.631-06:00Isn't the search space for nearest-neighbor as...Isn't the search space for nearest-neighbor assigning of size n choose k (which is < n^k), not n^{kd}, since each way to choose k unique points from n induces a candidate Voronoi partition? I don't see how the number of dimensions matters; there are only n choose k ways to pick the cluster centers. Or are you assuming that the center of a cluster need not be one of the n points?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6555947.post-76031071558587013082009-06-28T10:48:20.662-06:002009-06-28T10:48:20.662-06:00I'll be covering correlation clustering in a f...I'll be covering correlation clustering in a few posts :). Thanks for the comments: I'll update things shortly.Sureshhttp://www.blogger.com/profile/15898357513326041822noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-22079619047248078612009-06-28T07:10:28.372-06:002009-06-28T07:10:28.372-06:00Nice, One intriguing concept is Correlation Cluste...Nice, One intriguing concept is Correlation Clustering which captures nicely and directly the notion of pairwise hate and love among points... Sadly, the results knonw about this clsutering are not that exciting... --SAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-6555947.post-79079256989575422762009-06-28T05:01:10.490-06:002009-06-28T05:01:10.490-06:00Can you give a reference
for the Gonzalez 2-approi...Can you give a reference<br />for the Gonzalez 2-approimation<br />result? Thanks.<br /><br />Larry Wmehttp://www.blogger.com/profile/17047965327176583090noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-32785676808476035092009-06-28T03:28:13.723-06:002009-06-28T03:28:13.723-06:00Thanks for a good article.
There is one minor typ...Thanks for a good article.<br /><br />There is one minor typo:<br /><br />infer something about the distance between B and C<br />===><br />infer something about the distance<br />between A and C.Hongseok Yanghttp://www.blogger.com/profile/08010912896367131953noreply@blogger.com