Wiki
Clone wikibnpy-dev / demos / DemoIndex
Gaussian Toy Data Demos
Why initialization matters
Experiment showing how bnpy makes it easy to compare different initialization routines across many algorithm runs.
Births and merges inference for DP mixtures of Gaussians
Experiment showing how births and merges can find the ideal set of clusters, no matter how many we have initially.
Bag-of-words Toy Data Demos
Birth and merge inference for DP mixtures of Multinomials
Experiment showing how births and merges add and remove clusters to find the ideal set of bars.
Latent Dirichlet Allocation topic modeling with variational inference
Experiment shows the basics for training a topic model, comparing different number of topics.
Merge and delete moves with HDP topic models
Use merge and delete moves for topic models to identify the 10 true bars topics from initializations with many more.
Updated