Proper Losses for Learning with Example-Dependent Costs

by Alexander Hepburn, Ryan McConville, Raul Santos-Rodriguez, Jesus Cid-Sueiro and Dario Garcia-Garcia

This is a github repopository containing code needed to reproduce the results found in the paper.

Prerequesites

You will need : pytorch (if you don't have a GPU you need to remove all the .cuda() commands on the pytorch variables), torchvision, scikit-learn, CostCla, Pandas,

You will also need the files of the datasets used. For German Credit case study : https://archive.ics.uci.edu/ml/datasets/statlog+(german+credit+data)

For Music genre classificaiton: https://labrosa.ee.columbia.edu/millionsong/ http://www.ifs.tuwien.ac.at/mir/msd/ https://labrosa.ee.columbia.edu/millionsong/tasteprofile