Convergence of sequences of iterates of random-valued vector functions (Q4815087)

Given a probability space \((\Omega,\mathcal{A},P)\) and a closed subset of a separable Banach lattice \(E\), the author considers functions \(f:X\times\Omega\to X\) and their iterates \(f^n:X\times\Omega^{\mathbb{N}}\to X\) defined by \(f^1(x,\omega):=f(x,\omega_1)\) and \(f^{n+1}(x,\omega):=f(f^n(x,\omega),\omega_{n+1})\). Assume that \(f\) is random-valued, i.~e., measurable with respect to the \(\sigma\)-algebra \(\mathcal{B}\otimes\mathcal{A}\), where \(\mathcal{B}\) is the \(\sigma\)-algebra of Borel sets of \(X\), and that the expectation \(\mathbf{E}\| f^n(x,.)\|\) is finite for all \(x\) and \(n\). Then, assuming furthermore that \(m:X\to E\), \(m(x):=\mathbf{E} f^n(x,.)\), is continuous and that either \(m(x)\leq x\) for all \(x\) or \(m(x)\geq x\) for all \(x\), it is shown that \(m\circ\xi=\xi\), provided that \(E\) contains no subspaces isomorphic to \(c_0\) and that \((f^n(x_0,.))\) converges to \(\xi:\Omega^\mathbb{N}\to X\) in measure. If, additionally, \(E\) is an AL-space (\(\| x+y\|=\| x\|+\| y\|\) for all \(x,y\geq 0\) in \(E\)) it is also proved that the sequence \((f^n(x_0,.))\) converges a.~s. and in the \(L^1\)-norm to a unique fixed point of \(m\) provided that additionally \(m\) is a contraction. The author derives some consequences for \(E\)'s with special properties. (The reader should be aware of the fact that the author denotes the expectation and the Banach lattice with the same letter \(E\)!)