1: \begin{abstract}

2:         The Wang-Landau algorithm aims at sampling from a probability

3:         distribution, while penalizing some regions of the state space and

4:         favouring others. It is widely used, but its convergence properties are

5:         still unknown.  We show that for  some variations of the algorithm, the

6:         Wang-Landau algorithm reaches the so-called Flat Histogram criterion

7:         in finite time, and that this criterion can be never reached for other

8:         variations. The arguments are shown on an simple context -- compact

9:         spaces, density functions bounded from both sides-- for the sake of

10:         clarity, and could be extended to more general contexts.

11:     \end{abstract}