vlfeat库中的vl_kmeans函数

vl_kmeans支持不同的初始化和优化方法以及不同的聚类距离,具体参数如下:

Verbose::
%     Increase the verbosity level (may be specified multiple times).

 Distance::  默认[L2]
%     Use either L1 or L2 distance.

Initialization::   (初始聚类中心 
%     Use either random data points (RANDSEL) or k-means++ (PLUSPLUS) to initialize the centers.

Algorithm::  默认[LLOYD]        (聚类方法)

%     LLOYD, ELKAN, ANN中的一个

 NumRepetitions:: [1]
%     Number of time to restart k-means. The solution with minimal energy is returned.

以下选项调整ANN算法中用于ANN计算的KD-Tree森林

 NumTrees:: [3]
%     The number of trees int the randomized KD-Tree forest.

 MaxNumComparisons:: [100]

%     Maximum number of sample-to-center comparisons when searching for the closest center.搜索最近中心时的最大样本中心比较次数?

MaxNumIterations:: [100]

%      kmeans算法能够收敛的最大迭代次数


Example::
%     VL_KMEANS(X, 10, 'verbose', 'distance', 'l1', 'algorithm', 'elkan') clusters the data point X using 10 centers, l1 distance, and the Elkan's algorithm.


阅读更多
换一批

没有更多推荐了,返回首页