ml - Machine Learning in Python

This is a personal project whose goal is to concentrate all my Python code about Machine Learning.

You are completely free to use it anyway you want. However, it'd be cool if I was notified... just to know that this is useful for someone else =).

Of course, use it without any warranty... at your own risk. I can't guarantee complete documentation and stability. I make it stable and documented just enough for my own purposes.

Also, any contribution (new algorithms and, particularly, tests) are welcome. Send me a pull request or a patch and i'll be very happy to review it.

Currently, the following features are implemented:

• Supervised Learning:

  • Bayesian classifier based on Gaussian Discriminant Analysis
  • Multivariate Kernel Density Estimation
  • Bayesian classifier based on Kernel Density Estimation
  • kNN and Weighted-kNN classifiers
  • Rules for combination of classifiers (sum, product, median, min, max, majority) with single/multiple feature spaces
  • Stratified k-Fold cross validation
  • Some convenient functions to help evaluation

• Unsupervised Learning:

  • Partitioning Around Medoids (PAM) realization of the k-Medoids class of clustering algorithms
  • Some convenient functions to help partition clustering evaluation

• Statistics:

  • Efficient mode and frequency count functions
  • Multivariate normal PDFs

Dependencies

This project depends upon the following libraries:

• NumPy >= 1.6.0 (http://numpy.scipy.org/)
• SciPy >= 0.9.0 (http://www.scipy.org/)