Code written by Roland

• Hidden Markov Models in Python. Allows discrete and Gaussian observations and training on a single or on multiple observation sequences.
  
  
• Logistic regression in Python.
  
  
• Ultra-fast hebbian K-means algorithm in about 10 lines of code. Example application script.
  
  
• Gaussian mixture models in Python.
  
  
• A GPU implementation of a gated Boltzmann machine. Code written jointly with Josh Susskind. The module uses Volodymyr Mnih's cudamat package.
  
  
• A slightly older CPU implementation of a gated Boltzmann machine along with examples on this page.
  
  
• minimize.py is a Python port of Carl Rasmussen's matlab function minimize.m
  
  
• Code for computing kernel information embeddings. GPU-enabled through Tijmen Tieleman's gnumpy package.
  
  
• The gated softmax model is a very large mixture of logistic regressors with weight sharing. This implementation uses Volodymyr Mnih's cudamat package for training on GPUs.
  
  
• Gated auto-encoder using theano.
  
  
• I have been using Python for almost all of my research since about 2005.
  Here are some slides and code-snippets from a ten-minute mini-tutorial I gave to the machine learning group in Toronto in 2007 on switching from Matlab to Python.