Figure 1

The sliding window method. In each iteration step, the window W _i captures the set S _i of states in which the significant part of the probability mass is initially located (light gray), the set S_i+1of states that are reached after a time step (dark gray), as well as the states that are visited in between.