1: \begin{abstract}

2: We provide a  set of conditions which ensure the almost sure convergence of a class of simulated annealing algorithms on a bounded set $\setX\subset\mathbb{R}^d$ based on a time-varying Markov kernel. The class of algorithms considered in this work encompasses the one studied in \citet{Belisle1992} and \citet{Yang2000} as well as its derandomized version recently proposed by \citet{QMC-SA}. To the best of our knowledge, the results we derive 

3: are the first examples of almost sure convergence results for simulated annealing based on a time-varying kernel. In addition, the assumptions on the Markov kernel and on the cooling schedule have the advantage of being trivial to verify in practice.

4: 

5: 

6: \textit{Keywords:} Digital sequences; Global optimization; Simulated annealing

7: \end{abstract}

8: