jzyxn / ball-k-means

ball-k-means

• Ball k-means algorithms is described in detail in https://ieeexplore.ieee.org/document/9139397.

• the implementation of the ball k-means algorithm of the C++ version can be found in the "C++Version" file.

• the implementation of the ball k-means algorithm of the Python version can be found in the "PythonVersion" file.

• All data used in the paper is in the compressed file "data+centers(1).zip".

C++/Python version (C++/python 版本):

• the implementations of the ball k-means algorithm are "ball_k_means_Xf.cpp"/"ball_k_means_Xf.py" and "ball_k_means_Xd.cpp"/"ball_k_means_Xd.py", which are code for "float" and "double" versions respectively.

• the param "isRing" is used to switch the ring version and the no ring version of the algorithm.

• According to our experience, the "Xd" version can get more accurate results but the running time is slightly slower than "Xf"; the "Xf" version can reach the fastest running time, but low accuracy may result in many decimal places of data .

Requirements (环境要求)

Minimal installation requirements (C++) (最低需要安装要求):

• C++ compiler supporting C++11

• Linux operating system or Windows operating system

• Eigen 3 template library

Optional but recommended (C++) (可选安装要求但建议):

• BLAS implementation, we recommend this one: http://www.openblas.net/

• Intel MKL implementation, we recommend this one: https://software.intel.com/en-us/mkl

Installation requirements (Python) (安装要求):

• Only need to rely on the DLL files in the "PythonVersion" file.

Installation (C++) (安装)

• Eigen 3: In order to use Eigen, you just need to download and extract Eigen's source code: http://eigen.tuxfamily.org/index.php?title=Main_Page#Download

• ball_k_means_noRingVersion.cpp and ball_k_means_RingVersion.cpp both can be executed directly, only need to import Eigen library.

Using (用法)

C++ version (C++ 版本):

Step 1: call "ball_k_means" function. (调用"ball_k_means"函数)

Parameters (参数说明):

• dataset: clustering data in Matrix format in the Eigen library.

• centroids: initial center point data in matrix format in the Eigen library.

• isRing: bool type, optional parameters, switch the ring version and the no ring version of the algorithm. "true" means the current algorithm is a ring version, and "false" means the current algorithm is no ring version. The default is false.

• detail: bool type, optional parameters, "true" means output detailed information (including k value, distance calculation times, time, etc.), "false" means no detailed information is output. The default is false.

Output (输出说明):

• labels: labels of clustering data in matrix format in the Eigen library.

python version (python 版本):

Step 1: declare class "ball_k_means" and initialize algorithm. (声明"ball_k_means"类，算法初始化)

Parameters (参数说明):

• isRing: bool type, optional parameters, switch the ring version and the no ring version of the algorithm. "true" means the current algorithm is a ring version, and "false" means the current algorithm is no ring version. The default is false.

• detail: bool type, optional parameters, "true" means output detailed information (including k value, distance calculation times, time, etc.), "false" means no detailed information is output. The default is false.

Step 2: call "fit" function (调用"fit"函数)

Parameters (参数说明):

• dataset: absolute path of th csv file of clustering data.

• centroids: absolute path of th csv file of initial center point data.

Output (输出说明):

• labels: labels of clustering data in numpy matrix format.

Examples (示例):

Doesn't work? (有疑问？)

• Please contact Yong Zheng at [e-mail]