No label defined (Q3152606)

Let \((\Omega,{\mathfrak O},\mu)\) be a probability space, \({\mathfrak S}\) a sub-\(\sigma\)-field of \({\mathfrak O}\), \(E\) a separable Banach space, \(L^1_E({\mathfrak O})\) the space \(\mu\)-integrable \(E\)-valued functions and \(L^1_C({\mathfrak S})\) the space of \({\mathfrak S}\)-measurable functions in \(L^1_E({\mathfrak O})\) with values in a closed convex subset \(C\) of \(E\). The author studies the problem when every element of \(L^1_E({\mathfrak O})\) has a best approximation in \(L^1_C({\mathfrak S})\). Denote by \(\mu_C(f|{\mathfrak S})\) the set of best approximations of \(f\in L^1_E({\mathfrak O})\) in \(L^1_C({\mathfrak S})\). Under certain assumptions on \(C_n\) and \({\mathfrak S}\), the author gives information on \(\mu_{C_n}(f|{\mathfrak S}_n)\). For the entire collection see [Zbl 0993.00021].