1: \begin{abstract}

2: In this article, we present a greedy algorithm based on a tensor product decomposition, whose aim is to compute the global minimum of a strongly convex energy functional. We 

3: prove the convergence of our method provided that the gradient of the energy is Lipschitz on bounded sets. The main interest of this  

4: method is that it can be used for high-dimensional nonlinear convex problems. We illustrate this method on a prototypical example for uncertainty 

5: propagation on the obstacle problem.

6: \end{abstract}

7: