A noncompact Schauder fixed point theorem in random normed modules and its applications (Q7025791)

Let \((\Omega, \mathcal{F}, P)\) be a probability space with \(\sigma\)-field \(\mathcal{F}\) of subsets of \(\Omega\) and probability measure \(P\). Let \(K\) be the field of scalars (real or complex). Let \(L^0(\mathcal{F})\) be the algebra of (equivalence classes of) \(K\)-valued \(\mathcal{F}\)-measurable random variables \(\xi\). Let \(E\) be a linear space over \(K\). One says that \((E,\|\cdot\|)\) is a random normed module if \(E\) is a left module over \(L^0(\mathcal{F})\) and \(\|\cdot\|: E\to \{\xi\in L^0(\mathcal{F}): \xi\geq 0\}\) behaves like a norm on \(E\): \(\|\xi x\| = |\xi| \|x\|\); \(\|x+y\| \leq \|x\| + \|y\|\); \(\|x\|=0 \Rightarrow x=0\in E\) for all \(x,y\in E\), \(\xi\in L^0(\mathcal{F})\). It should be noted that \(\mathcal{F} = \{\emptyset, \Omega\}\) leads to an ordinary normed space. There are various possible topologies for \(E\). However, the most useful one is the \((\epsilon,\lambda)\)-topology generated by the local base at \(0\in E\) consisting of neighborhoods \(\{x\in E: P[ \|x\|<\epsilon ] > 1-\lambda\}\), \(\epsilon>0\), \(0<\lambda<1\).\N\NA subset \(G\subset E\) is: (a) \(L^0\)-convex if \(\xi x+ (1-\xi)y \in G\) whenever \(x,y\in G\), \(\xi\in L^0(\mathcal{F})\), \(0\leq \xi \leq 1\); (b) \(\sigma\)-stable if for every sequence \((x_n)\) in \( G\) and every countable partition \(A_n\in\mathcal{F}\) of \(\Omega\), there exists \(x\in G\) such that \(I_{A_n} x = I_{A_n} x_n\) for all \(n\), where \(I_{A}\) stands for (the equivalence class of) the characterisic function of a set \(A\); (c) random sequentially compact if every sequence \((x_n)\) in \( G\) admits a random subsequence \((x_{n_k})\) converging in the \((\epsilon,\lambda)\)-topology to some \(x\in G\). A random subsequence of \((x_n)\) is \((x_{n_k}) = (\sum_{l=1}^{\infty} I_{\{n_k=l\}} x_l ) \), where \(n_k\) are positive integer-valued random variables on \((\Omega,\mathcal{F},P)\) and \(n_k(\omega) < n_{k+1}(\omega)\) for \(\omega\in \Omega\) and all \(k\). A map \(T:G\to G\), \(G\subset E\), is: (d) random sequentially continuous at \(x_0\in G\) if \(G\) is \(\sigma\)-stable and for every sequence \((x_n)\) in \( G\) that converges to \(x_0\) in the \((\epsilon,\lambda)\)-topology, there exists a random subsequence \((x_{n_k})\) such that \((T(x_{n_k}))\) converges to \(T(x_0)\) in the \((\epsilon,\lambda)\)-topology; (e) \(\sigma\)-stable if \(G\) is \(\sigma\)-stable and \(T(\sum_{n=1}^{\infty} I_{A_n} x_n) = \sum_{n=1}^{\infty} I_{A_n} T(x_n)\) for every sequence \((x_n)\) in \( G\) and every countable partition \(A_n\in\mathcal{F}\) of \(\Omega\).\N\NThe authors prove that every \(\sigma\)-stable, random sequentially continuous map \(T:G\to G\) acting on a random sequentially compact and \(L^0\)-convex subset \(G\subset E\) of a random normed module \(E\) admits a fixed point. Alongside the above randomized Schauder fixed point principle, the authors prove a conditional version of the Nash equilibrium theorem. The study is motivated by the lack of compactness of some sets which happen to be random sequentially compact. Finally, the authors point out that one cannot automatically transfer their considerations to random locally convex modules.