| Who | When |
Messages | |
|
|
|
Dave Bradley
|
1
|
 |
|
04-03-2006 08:48 AM ET (US)
|
|
Please post your thoughts on the effectiveness of non-linear dimensionality reduction here.
|
| Tomasz Malisiewicz
|
2
|
|
|
04-29-2006 01:53 AM ET (US)
|
|
A really great resource for non-linear dimensionality reduction techniques are the Lawrence Saul video lectures (4 of them; each one is 1 hr) found at: http://www.oid.ucla.edu/Webcast/ipam/Enjoy! |
| Mohit Gupta
|
3
|
|
|
04-29-2006 06:46 PM ET (US)
|
|
These methods work great on the examples shown in the paper, and are indeed powerful Non-Linear DImensionality reduction tools.
However, it would be interesting to see how they deal with noise -- for example, if there is a small, thin bridge connecting two adjacent layers of the swiss roll... would they go completely berserk? Or will they still be able to do something intelligent? I somehow suspect the former, specially for LLE.
|
| Mohit Gupta
|
4
|
|
|
04-29-2006 06:58 PM ET (US)
|
|
How do LLE and IsoMap compare in terms of running time? IsoMap is O(N^2 logn); I am not sure about LLE?
|
| Alyosha Efros
|
5
|
|
|
05-01-2006 05:27 AM ET (US)
|
|
|
|
|
6
|
|
|
07-22-2006 02:32 AM ET (US)
|
|
Deleted by topic administrator 07-22-2006 10:25 AM
|
