pschulam / kernels

Various kernels (gibbs, slice, MH) for datamicroscopes

microscopes-kernels

Contains various ergodic MCMC kernels (gibbs, slice, metropolis-hastings). Used to do inference on the non-parametric models found in both the mixturemodel and the irm project. More models (such as HDP-HMM and HDP-LDA) will be added in the future. Stay tuned!

Installation

Follow the instructions in the common project for setting up your Anaconda environment and adding the necessary binstar channels. Then run

$ conda install microscopes-kernels

Examples

Here are a few ipython notebooks. More examples can be found in the examples directory.

• 2D Gaussian mixture model
• Digit completion with MNIST data
• Enron e-mail corpus clustering
• Cancer genomics clustering