Description:

A collection of python code to perform research in optimization. The aim is to provide reusable components that can be quickly applied to machine learning problems.

Dependencies:

• python 2.5+
• cvxopt 1.0+ (for solving linear and quadratic programs)
• pythongrid (for using a cluster)
• cython 0.14.1 (for speeding up kernel computations)

Description

The projects currently prototyped:

• Ellipsoidal multiple instance learning
• Contextual bandits upper confidence bound algorithm (using GP)
• learning the output kernel using block coordinate descent
• difference of convex functions algorithms for sparse classfication

Ellipsoidal Multiple Instance Learning

• The code for eMIL is contained in mil.mi_classifier_emil
• emil_demo.py gives usage examples
• mil.mi_data provides containers for multiple instance learning data

Learning the output kernel using block coordinate descent

• The workhorse is kernelopt.py, which implements a framework for minimizing the invex function for learning the kernel on outputs.
• multiclass_demo.py gives usage examples.

Changes to previous version:

• minor bugfix

BibTeX Entry: Download

Supported Operating Systems: Agnostic

Data Formats: Matlab, Hdf, Numpy, Json

Tags: Kernel Learning

Archive: download here

Other available revisons

Version	Changelog	Date
0.3.1	minor bugfix	May 2, 2013, 10:46:11
0.3	Included code for Gaussian Process Contextual Bandits Implemented Ellipsoidal Multiple Instance Learning difference of convex functions algorithms for sparse classfication	April 4, 2013, 12:48:56
0.2.1	Fixed missing multiclass module. Slycot sources no longer distributed, using github project instead.	July 21, 2011, 20:39:12
0.2	Initial Announcement on mloss.org.	May 1, 2011, 15:30:29

Comments

No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.