Sean Noran(snoran@umass.edu)

In this work, we demonstrate that Multi-Instance Learning (MIL) can be applied to various time-domain datasets, in order to reduce the need for fine-grained labels. We examine various MIL algorithms.

TODO: Add installation instructions (see misvm, miforest)

...

or by running the setup file manually

git clone [ url for MIL ]
cd MIL
python setup.py install

misvm depends on the following packages. (see TODO: add link to misvm)

• NumPy: Library for efficient matrix math in Python
• SciPy: Library for more MATLAB-like functionality
• CVXOPT: Efficient convex (including quadratic program) optimization

...