Stochastic Embedding revisited: A modern interpretation

There is a very extensive literature on various aspects of the central Bias-Variance trade-off in linear system identification. In the 80´s and 90´s the focus was on bias characterization, model error models and Stochastic Embedding. Recently, there has been a new interest in Bayesian or kernel methods. This paper puts part of this literature into perspective by giving a modern interpretation of the Stochastic Embedding approach.