Skip to content

Kmcuda

kmcuda is a large scale K-means and K-nn implementation on NVIDIA GPU / CUDA.[1] The implementation is based on Ding, et al.[2]

Installed Version

The latest main branch (as of 2019-11-22) is installed on the GPU nodes. It is installed in the python37 conda environment, using the python/anaconda3 modulefile.

$ module load python/anaconda3 $ conda activate python37 (python37) $

See Anaconda#Conda Environments for infomation on using conda environments.

References

[1] kmcuda GitHub repository

[2] Yinyang K-Means: A Drop-In Replacement of the Classic K-Means with Consistent Speedup (PDF)