LFD Book Forum

LFD Book Forum (http://book.caltech.edu/bookforum/index.php)
-   Homework 8 (http://book.caltech.edu/bookforum/forumdisplay.php?f=137)
-   -   K-means clustering for RBF centers (http://book.caltech.edu/bookforum/showthread.php?t=613)

leonidr 06-02-2012 04:27 PM

K-means clustering for RBF centers
 
Why do we disregard the labels when we pick the centers?

I know that the first step is an example of unsupervised learning. But isn't the overall goal classification? And we measure the error of that classification based off of the labels. So, in practice, why would we disregard them in a typical supervised problem?

I know that one difficulty that immediately comes to mind is how do we combine (and/or map) the x and y into a space where the clustering can be performed easily; where the distance metric between two points weighs the x and the y appropriately (an open question I guess). But perhaps a naive approach such as appending the label (+-1) would have some benefit?


My apologies for asking a question outside the realm of the book, but I thought that it was relevant for the class. Any insight would be interesting.

dudefromdayton 06-03-2012 07:25 AM

Re: K-means clustering for RBF centers
 
Quick but very incomplete answer: by disregarding the labels, we don't have to charge for looking at the data. That is, we aren't snooping. This greatly reduces the dimensionality of the hypothesis set that we might ultimately formulate an answer from later. But I think there are other reasons, and I have to rush to cover an obligation.


All times are GMT -7. The time now is 10:25 PM.

Powered by vBulletin® Version 3.8.3
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.
The contents of this forum are to be used ONLY by readers of the Learning From Data book by Yaser S. Abu-Mostafa, Malik Magdon-Ismail, and Hsuan-Tien Lin, and participants in the Learning From Data MOOC by Yaser S. Abu-Mostafa. No part of these contents is to be communicated or made accessible to ANY other person or entity.