tracking segmentation

Comment

Comment by David Stasiuk on April 23, 2014 at 2:26pm

Thanks Robert! The whole generative side of the project completely relied on OctopusE...this is just some descriptive data viz on the second part. This is really all just a way of looking at several instances of a k-means algorithm run on the same data set. I used 18 descriptive variables to feed the algorithm, and then set it to create outputs from 2 clusters to 80...so it isn't determining the number itself at all. That's for what's next...here I'm trying to visualize how constituent elements move from iteration to iteration as more clusters are introduced...each cluster is sorted in each iteration according to the weighted "location" of the instances from the previous iteration, so pure black/white in the 2 cluster pass gets diluted as they mix together, and more colors get embedded to reflect the nuances that come into play when you increase the number of clusters...otherwise it just turns grey.

Comment by Robert Vier on April 23, 2014 at 2:09pm

nice graphics! which algorithm do you use? is it determining the number of clusters itself?