These can be supported both via cluster and reduce:
Cluster: behaves similarly to GaussianMixtureModel (estimate how much of each factor is reflected by/in each observation)
Reduce: to reduce the data to n-dimensions, fit LDA/NMF with n components
We could play a similar trick with Gaussian mixture models-- to use a Gaussian mixture model to reduce the data to n dimensions, we could fit a GMM with n components and then use the membership mixing labels as the coordinates.
We should add support for LDA and NMF: http://scikit-learn.org/stable/modules/classes.html#module-sklearn.decomposition
These can be supported both via
clusterandreduce:Cluster: behaves similarly to GaussianMixtureModel (estimate how much of each factor is reflected by/in each observation) Reduce: to reduce the data to n-dimensions, fit LDA/NMF with n components
Some sample code may be found here: https://github.com/ContextLab/storytelling-with-data/blob/master/data-stories/twitter-finance/twitter-finance.ipynb
We could play a similar trick with Gaussian mixture models-- to use a Gaussian mixture model to reduce the data to n dimensions, we could fit a GMM with n components and then use the membership mixing labels as the coordinates.