Initialization of centroid for kmeans++ algorithm

5 次查看（过去 30 天）
AS
AS 2020-9-20
评论： Image Analyst 2020-10-1
I have applied simple kmeans algorithm for clustering of the datasets, which has dimension 1800 by 3. Now instead of kmeans, I want to initialize the centroid by using kmeans++ algorithm. For simple kmeans I have initialized the centres as per algorithm like Arbitrarily choose an initial k centers C = {c1, c2, · · · , ck}.
But for kmeans++ the initialization procedure is like that
1a. Take one center c1, chosen uniformly at random from X .
1b. Take a new center ci , choosing x ∈ X with probability (D(x)^2) /sum(D(x)^2) .
1c. Repeat Step 1b. until we have taken k centers altogether.
In particular, let D(x) denote the shortest distance from a data point to the closest center. “D2 weighting”
But, I am not understanding the “D2 weighting”.
How to calculate “D2 weighting” value for initialization.
Please provide me any advice regarding this.

请先登录，再回答此问题。

回答（1 个）

Image Analyst
Image Analyst 2020-9-20
I believe you can use the 'Replicates' and 'Start' options of kmeans().

类别

在 Help Center 和 File Exchange 中查找有关 Statistics and Machine Learning Toolbox 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by