1: \begin{abstract}

2: We consider an incremental approximation method for solving variational problems in infinite-dimensional separable Hilbert spaces, where in each step a randomly and independently selected subproblem from an infinite collection of subproblems is solved. We show that convergence rates for the expectation of the squared error can be guaranteed under weaker conditions than previously established in \cite{GrOs2016}. 

3: 

4: 

5: \keywords{infinite space splitting \and subspace correction \and multiplicative Schwarz \and block coordinate descent \and    greedy \and randomized \and convergence rates \and online learning}

6: \subclass{65F10 \and 65N22\and 49M27}

7: \end{abstract}