Milk is a machine learning toolkit in Python.

Its focus is on supervised classification with several classifiers available: SVMs (based on libsvm), k-NN, random forests, decision trees. It also performs feature selection. These classifiers can be combined in many ways to form different classification systems.

For unsupervised learning, milk supports k-means clustering and affinity propagation.

Milk is flexible about its inputs. It optimised for numpy arrays, but can often handle anything (for example, for SVMs, you can use any dataype and any kernel and it does the right thing).

There is a strong emphasis on speed and low memory usage. Therefore, most of the performance sensitive code is in C++. This is behind Python-based interfaces for convenience.

To learn more, check the docs at https://packages.python.org/milk/ or the code demos included with the source at milk/demos/.

Here is how to test how well you can classify some features,labels data, measured by cross-validation:

import numpy as np
import milk
features = np.random.rand(100,10) # 2d array of features: 100 examples of 10 features each
labels = np.zeros(100)
features[50:] += .5
labels[50:] = 1
confusion_matrix, names = milk.nfoldcrossvalidation(features, labels)
print 'Accuracy:', confusion_matrix.trace()/float(confusion_matrix.sum())

If want to use a classifier, you instanciate a learner object and call its train() method:

import numpy as np
import milk
features = np.random.rand(100,10)
labels = np.zeros(100)
features[50:] += .5
labels[50:] = 1
learner = milk.defaultclassifier()
model = learner.train(features, labels)

# Now you can use the model on new examples:
example = np.random.rand(10)
print model.apply(example)
example2 = np.random.rand(10)
example2 += .5
print model.apply(example2)

There are several classification methods in the package, but they all use the same interface: train() returns a model object, which has an apply() method to execute on new instances.

License: MIT

Author: Luis Pedro Coelho (with code from LibSVM and scikits.learn)

API Documentation: https://packages.python.org/milk/

Mailing List: https://groups.google.com/group/milk-users

• SVMs. Using the libsvm solver with a pythonesque wrapper around it.
• K-means using as little memory as possible. It can cluster millions of instances efficiently.
• Random forests
• Self organising maps
• Stepwise Discriminant Analysis for feature selection.
• Non-negative matrix factorisation
• Affinity propagation

The ChangeLog file contains a more complete history.

• Add folds argument to nfoldcrossvalidation
• Add assign_centroid function in milk.unsupervised.nfoldcrossvalidation
• Improve speed of k-nearest neighbour (10x on scikits-learn benchmark)
• Improve kmeans on newer numpy (works for larger datasets too)
• Faster kmeans by coding centroid recalculation in C++
• Fix gridminize for low count labels
• Fix bug with non-integer labels for tree learning

• Fix compilation on Windows

• Logistic regression
• Source demos included (in source and documentation)
• Add cluster agreement metrics
• Fix nfoldcrossvalidation bug when using origins

• Unsupervised (1-class) kernel density modeling
• Fix for when SDA returns empty
• weights option to some learners
• stump learner
• Adaboost (result of above changes)

• Fixes for 64 bit machines
• Functions in measures.py all have same interface now.