Implementation of Multimodal-GMM with Gibbs sampling algorithm.
Allows clustering of data sampled from two or more different multivariate normal distributions
- The first step is to create the observation data using make_data.py. Then, create data1.txt and data2.txt. true_label.txt is the label data for calculating ARI.
- After that, you can use mgmm.py to run the clustering.
The image below shows the actual generated observables for the two modalities.(The cluster numbers for the two data points are the same)
The image below shows the actual ARI measured by mgmm.py, where a value close to 1 means high cluster performance and a value close to 0 means low cluster performance.
Other examples of implementations using GMM are as follows