Welcome to MLPython’s documentation!

MLPython is a library for organizing machine learning research. There are two reasons why you might be interested in MLPython:

1. to rapidly gain access to many datasets and learning algorithms or
2. to organize your own machine learning research into a consistent, flexible and simple framework.

I developed MLPython to support my own research. I wanted a system that would allow me to quickly experiment with new research ideas. Having found it to deliver well on this account, I decided to share MLPython with others. I hope you enjoy it!

• Installation Instructions: Details on how to download and install MLPython.
• Tutorial: How to use MLPython: A walk-through over diffrent examples of how to use MLPython.
• Extending MLPython: Guidelines for contributing to MLPython and extending it.
• Library Documentation: Detailed documentation, extracted from the source code docstrings.

Here’s a quick peek at what MLPython let’s you do:

import numpy as np
import mlpython.datasets.store as dataset_store
from mlpython.learners.classification import NNet
from mlpython.learners.features import RBM

dataset_store.download('mnist')
trainset,validset,testset = dataset_store.get_classification_problem('mnist')

nnet = NNet(n_stages=10)
nnet.train(trainset)

outputs,costs = nnet.test(testset)
print 'Classification error on test set is',np.mean(costs,axis=0)[0]

As is probably obvious, this code snippet downloads the MNIST dataset, trains a feed-forward neural network for 10 iterations on it and finally computes the average classification error on a test set.

Indices and tables

• Index
• Module Index
• Search Page