Presentation is loading. Please wait.

Presentation is loading. Please wait.

Random Swap EM algorithm for GMM and Image Segmentation

Similar presentations


Presentation on theme: "Random Swap EM algorithm for GMM and Image Segmentation"— Presentation transcript:

1 Random Swap EM algorithm for GMM and Image Segmentation
Qinpei Zhao, Ville Hautamäki, Ismo Kärkkäinen, Pasi Fränti Speech & Image Processing Unit Department of Computer Science, University of Joensuu Box 111, Fin Joensuu FINLAND

2 Outline Background & Status RS-EM Application

3 Background: Mixture Model

4 Background: EM algorithm
E-step (Expectation): M-step (Maximization): Iterate E,M step until convergence α- mixing coefficient Θ- model parameters, eg. {μ,∑}

5 Local Maxima Let’s describe it as mountain climbing……

6 2160m 3099m 600km

7

8 Initialization Effect
Initialization and Result(1) Initialization and Result(2)

9 The situation of local maxima trap
Sub-optimal Example The situation of local maxima trap

10 Status Standard EM for Mixture Models(1977)
Deterministic Annealing EM (DAEM) (1998) Split-Merge EM (SMEM) (2000) Greedy EM (2002) RS-EM coming…

11 Outline Background & Status RS-EM (Random Swap) Application

12 RSEM: Motivations Random manner
Prevent from staying near the unstable or hyperbolic fixed points of EM. Prevent from its stable fixed points corresponding to insignificant local maxima of the likelihood function Avoid the slow convergence of EM algorithm Less sensitive to its initialization

13 Formulas SMEM Greedy EM RSEM

14 Random Swap EM After EM After EM After Swap

15 Comparisons(1)

16 Comparisons(2) Q1 Q2 S1 S4

17 Outline Background & Status RS-EM Application

18 Application Image Segmentation Color Quantization Image Retrieval ……

19 Conclusion Thanks!☺ Introduce Randomization into algorithm
Performs better Without heavy time complexity Wider applications Thanks!☺


Download ppt "Random Swap EM algorithm for GMM and Image Segmentation"

Similar presentations


Ads by Google