implement k-means++ initialization - Githubissues

SOCR / SOCRAT-issues

Project management for SOCRAT

2 stars 0 forks source link

implement k-means++ initialization #165

Closed alxndrkalinin closed 6 years ago

alxndrkalinin commented 7 years ago

k-means++ is an initialization algorithm for choosing the initial values (or "seeds") for the k-means

[ ] read how algorithm works
[ ] create new service kMeansPlusPlus and inject it into ClusterKMeans
[ ] service should provide method init that implements k-means++ algorithm
[ ] add k-means++ to the list of initialization methods by linking init method here
[ ] test that algorithm works using Neuroimaging study of 27 AD, MCI, NC subjects dataset and make it default init method
[ ] create pull request to SOCR/SOCRAT:dev

Note: for algorithm implementation you can use scikit-learn Python code as an example and ClusterMetrics service to calculate distances