Master	Master	Develop	Develop

A lightweight toolkit to help you on common Machine Learning tasks. Implement two interfaces and start experimenting.

Warning: this project is still in a very alpha stage.

To install the package locally:

$ pip install -r requirements.txt
$ make install

Using:

$ mlpal --help

$ make uninstall

For mlpal to work, you must define a setup module, that will contain two classes:

• DataSource
• LearningSpec

A working example of setup file is the tests/dummy_setup.py.

The tasks that are already supported:

• train
• search
• benchmark
• learning_curves
• plot_pca
• misclassified

Most command line parameters can have their default value defined in the config.yaml file.

Random acts of kindness are more than welcome. Besides the Github issues, known smells in the code can be listed via:

$ make todos

You may also want to take a look at these projects:

• SKLL: http://skll.readthedocs.org
• nltk-trainer: http://nltk-trainer.readthedocs.org