Open fch-aa opened 8 years ago
Hi there,
Thanks a lot for bringing the issue! It would indeed be very nice to have this supported in the framework, but I am afraid right now it is not exactly a top priority in the project.
In case someone would like to give this a shot, the references for the use of Triangle Inequality for accelerating k-means can be found in http://users.cecs.anu.edu.au/~daa/courses/GSAC6017/kmeansicml03.pdf.
By the way, another related issue is the creation of a mini-batch version of k-means. In my experience, such version would be even faster than a possible accelerated-triangle-inequality version. A reference for batch k-means can be found in http://www.eecs.tufts.edu/~dsculley/papers/fastkmeans.pdf.
Regards, Cesar
I see that it was mentioned in the "Clustering/KMeans/KMeans.cs" file, any plans on working on this?