In this video, we see 100 data points distributed over 25 nodes. The data is modelled bij a Mixture of Gaussians with 5 mixtures (each represented by a colour). The ellipses show the current estimation of a node of the covariance for each cluster. We see that within a small communication steps the local estimates converge to the correct estimates. The parameters are initialised with a distributed implementation of the K-Means algorithm.
Link to this comment:
All Comments (0)