(copy from lect. 3) |
|||
| (One intermediate revision by the same user not shown) | |||
| Line 1: | Line 1: | ||
| + | [[Category:ECE662]] | ||
| + | [[Category:decision theory]] | ||
| + | [[Category:lecture notes]] | ||
| + | [[Category:pattern recognition]] | ||
| + | [[Category:slecture]] | ||
| + | |||
| + | =The Curse of Dimensionality= | ||
| + | from [[Lecture_2_-_Decision_Hypersurfaces_OldKiwi|Lecture 2, ECE662, Spring 2010]] | ||
| + | ---- | ||
Refers to the problem caused by exponential growth of hypervolume as a function of dimensionality. This term was coined by Richard Bellman in 1961. | Refers to the problem caused by exponential growth of hypervolume as a function of dimensionality. This term was coined by Richard Bellman in 1961. | ||
As stated in [[Lecture 3 - Bayes classification_Old Kiwi]], | As stated in [[Lecture 3 - Bayes classification_Old Kiwi]], | ||
The curse of dimensionality starts at d>17-23. There are no clusters or groupings of data points when d>17. In practice each point turns to be a cluster on its own and as a result this explodes into a high dimensional feature vectors which are impossible to handle in computation. | The curse of dimensionality starts at d>17-23. There are no clusters or groupings of data points when d>17. In practice each point turns to be a cluster on its own and as a result this explodes into a high dimensional feature vectors which are impossible to handle in computation. | ||
| + | ---- | ||
| + | [[Lecture_2_-_Decision_Hypersurfaces_OldKiwi|Back to Lecture 2, ECE662, Spring 2010]] | ||
Latest revision as of 09:57, 10 June 2013
The Curse of Dimensionality
from Lecture 2, ECE662, Spring 2010
Refers to the problem caused by exponential growth of hypervolume as a function of dimensionality. This term was coined by Richard Bellman in 1961.
As stated in Lecture 3 - Bayes classification_Old Kiwi, The curse of dimensionality starts at d>17-23. There are no clusters or groupings of data points when d>17. In practice each point turns to be a cluster on its own and as a result this explodes into a high dimensional feature vectors which are impossible to handle in computation.
