Validation in Statistics and Machine Learning - Abstract

Braun, Mikio L.

Machine Learning Open Source Software and Benchmark Repository

While everyone agrees that research should be reproducible, actually reproducing results from papers in statistics and machine learning can be difficult in practice. Important implementation details may be missing, and the data migth not be available. In this talk, I present our initiatives "machine learning open source software" (MLOSS) and "machine learning benchmark repository" (MLDATA) which aim at providing an open platform for exchanging software and data sets, and discuss the experiences we have made in trying to build a community around them, and the lessons we have learned.