tag:blogger.com,1999:blog-6555947.post7783488448402466968..comments2024-03-14T01:32:43.610-06:00Comments on The Geomblog: ECML-PKDD conference report: Part ISuresh Venkatasubramanianhttp://www.blogger.com/profile/15898357513326041822noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-6555947.post-64175844766795483582011-10-31T20:10:42.410-06:002011-10-31T20:10:42.410-06:00Very interesting post about clustering, I liked a ...Very interesting post about clustering, I liked a lot the idea of fractals or subspace clusters. <br /><br />I think that the main problem with clustering methods is that in order to make the best possible grouping with a certain method, you need to, understand first the data you are working with, because in that way you have much more chances than when this data set you are grouping is not reaching satisfactory results (or they don't even make any logical seance) of having correct intuitions of what is possible to improve, or which areas/concepts might be causing this problems. This is specially important and difficult while using a certain amount of metrics that where created by you and also other researchers, which might be very difficult to have a decent idea of what kind of information can that summary of metrics derive into<br /><br />Also the method internals must be clearly understood, along with implications of parameters variations in the way the data is processed.Mariana Sofferhttps://www.blogger.com/profile/13351209522681966230noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-74996281339284980522011-09-07T12:50:21.253-06:002011-09-07T12:50:21.253-06:00In non-theory conferences, the "PC" usua...In non-theory conferences, the "PC" usually includes people you'd normally think of as external reviewers for a theory conference. You get the tag of being on the PC, which is why it's easy to pile up the PC memberships, but the numbers are similar if you look at the external reviewer count for STOC/FOCS. <br /><br />PC members can submit papers as well, and usually only review 8-10 papers. It's the area chairs/vice chairs who act like a theory PC (sort of)Suresh Venkatasubramanianhttps://www.blogger.com/profile/15898357513326041822noreply@blogger.comtag:blogger.com,1999:blog-6555947.post-7100210424007636192011-09-07T12:47:44.702-06:002011-09-07T12:47:44.702-06:00The size of the PC seems to exceed the number of p...The size of the PC seems to exceed the number of papers!Chandranoreply@blogger.com