Memory loss for time-dependent dynamical systems (Q843062)

Let \(p_t\) denote the probability density with respect to a fixed reference measure \(m\) at time \(t\geq 0\), and \(p_0\) its initial density. A dynamical system evolving with \(p_t\) is said to \textit{loose its memory} in statistical sense with respect to \(m\) iff, for any two initial densities \(p_0\) and \(\tilde{p}_0\), \[ \lim_{t \to +\infty} \int_E |p_t - \tilde{p}_t | dm = 0 . \] Moreover, it has \textit{exponential loss of memory} in statistical sense with respect to \(m\) iff \[ \exists c, \alpha>0 \; \forall p_0, \tilde{p}_0 \; \forall t \geq 0 \; : \quad \int_E |p_t - \tilde{p}_t | dm \leq c \exp (-\alpha t) . \] The authors prove exponential loss of memory in that sense for time-dependent dynamical systems defined by expanding and piecewise expanding maps with slowly varying parameters (iterated function systems).