Review Note
Last Update: 02/01/2024 01:45 PM
Current Deck: Computer Vision
Published
Fields:
Vorderseite
Segmentation: Mixture of Gaussians
Rückseite
Assumption:
Assume data is generated by k weighted d-dimensional Gaussians. Parameters:
The likelihood of observing x is then:
(the a_i sum to 1)
Advantages:
+ is probabilistic approach and can detect outliers, aswell as generate new points
+ compact to store in O(k * (d^2 + d))
- need to know k in advance
- good initialization needed
Assume data is generated by k weighted d-dimensional Gaussians. Parameters:
The likelihood of observing x is then:
(the a_i sum to 1)
Advantages:
+ is probabilistic approach and can detect outliers, aswell as generate new points
+ compact to store in O(k * (d^2 + d))
- need to know k in advance
- good initialization needed
Suggested Changes:
Deck Changes (Suggestion to move the Note to the following Deck):
Field Changes:
Tag Changes: