An information geometrical view of stationary subspace analysis

Stationary Subspace Analysis (SSA) [3] is an unsupervised learning method that finds subspaces in which data distributions stay invariant over time. It has been shown to be very useful for studying non-stationarities in various applications [5,10,4,9]. In this paper, we present the first SSA algorithm based on a full generative model of the data. This new derivation relates SSA to previous work on finding interesting subspaces from high-dimensional data in a similar way as the three easy routes to independent component analysis [6], and provides an information geometric view.