Skip to main content
18 events
when toggle format what by license comment
Sep 22, 2021 at 23:23 history edited ttnphns CC BY-SA 4.0
added 2 characters in body
Sep 22, 2021 at 22:24 comment added ttnphns @Ben, I've added a footnote.
Sep 22, 2021 at 22:23 history edited ttnphns CC BY-SA 4.0
added 479 characters in body
Sep 22, 2021 at 20:11 comment added Ben @ttnphns, it's still not clear. Say I run k-means three times on a 2-d dataset, each time generating two centers: trial1:{(1.5, 0.1), (2.3, 1.1)}, trail2:{(0.4, 0.1), (3.1, 1.7)}, trial3:{(0.9, 1.5), (2.0, 2.0)}. What are the "average" centers here?
Sep 22, 2021 at 19:58 comment added ttnphns @Ben, as usual. Average their coordinates by each of the features.
Sep 22, 2021 at 16:00 comment added Ben How exactly does one "average the cluster centres of those runs"?
Nov 6, 2019 at 18:06 comment added ttnphns @camillejr, please start by checking this Q: stats.stackexchange.com/q/418427/3277.
Nov 6, 2019 at 13:10 comment added kamilazdybal @ttnphns thanks for helpful answer! I've a question about the sensitivity to object order - do you perhaps know any paper when this was showed/studied?
Mar 22, 2019 at 7:55 history edited ttnphns CC BY-SA 4.0
added 168 characters in body
Oct 18, 2018 at 12:09 comment added Herman Toothrot @ttnphns how do determine quantitatively that variables have "quite different variances"?
Feb 23, 2017 at 7:19 comment added ttnphns @StudentT, I've added a footnote for that. Thank you.
Feb 23, 2017 at 7:19 history edited ttnphns CC BY-SA 3.0
added 544 characters in body
Feb 23, 2017 at 5:39 comment added SmallChess How would k-means be sensitive to ordering?
Dec 13, 2016 at 2:50 history edited ttnphns CC BY-SA 3.0
pic added
Jan 30, 2012 at 16:08 comment added pedrosaurio The randomizing, re-running, averaging and final run is a very good advice. Thanks
Jan 30, 2012 at 16:02 vote accept pedrosaurio
Jan 17, 2012 at 11:33 history edited ttnphns CC BY-SA 3.0
added 344 characters in body
Jan 17, 2012 at 11:23 history answered ttnphns CC BY-SA 3.0