Stationary Subspace Analysis

Non-stationarities are an ubiquitous phenomenon in real-world data, yet they challenge standard Machine Learning methods: if training and test distributions differ we cannot, in principle, gen- eralis