mloss.org OptWokhttp://mloss.orgUpdates and additions to OptWokenThu, 02 May 2013 10:46:11 -0000OptWok 0.3.1http://mloss.org/software/view/312/<html><p>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. </p> <h1>Dependencies:</h1> <ul> <li> python 2.5+ </li> <li> cvxopt 1.0+ (for solving linear and quadratic programs) </li> <li> pythongrid (for using a cluster) </li> <li> cython 0.14.1 (for speeding up kernel computations) </li> </ul> <h1>Description</h1> <p>The projects currently prototyped: </p> <ul> <li> Ellipsoidal multiple instance learning </li> <li> Contextual bandits upper confidence bound algorithm (using GP) </li> <li> learning the output kernel using block coordinate descent </li> <li> difference of convex functions algorithms for sparse classfication </li> </ul> <h2>Ellipsoidal Multiple Instance Learning</h2> <ul> <li> The code for eMIL is contained in mil.mi_classifier_emil </li> <li> emil_demo.py gives usage examples </li> <li> mil.mi_data provides containers for multiple instance learning data </li> </ul> <h2>Learning the output kernel using block coordinate descent</h2> <ul> <li> The workhorse is kernelopt.py, which implements a framework for minimizing the invex function for learning the kernel on outputs. </li> <li> multiclass_demo.py gives usage examples. </li> </ul></html>Cheng Soon Ong, Gabriel KrummenacherThu, 02 May 2013 10:46:11 -0000http://mloss.org/software/rss/comments/312http://mloss.org/software/view/312/kernel learning