1: \begin{abstract}

2:   % Many statistical models in Bayesian analysis assume some constraints

3:   % on the parameters to infer.

4:  This paper presents a detailed theoretical analysis of the Langevin Monte Carlo sampling algorithm recently introduced in

5:   \cite{durmusprox:2016} when applied to log-concave probability

6:   distributions that are restricted to a convex body $\convSet$. This method relies on

7:   a regularisation procedure involving the Moreau-Yosida envelope of the

8:   indicator function associated with $\convSet$. Explicit

9:   convergence bounds in total variation norm and in Wasserstein

10:   distance of order $1$ are established. In particular, we show that

11:   the complexity of this algorithm given a first order oracle is

12:   polynomial in the dimension of the state space. Finally, some

13:   numerical experiments are presented to compare our method with

14:   competing MCMC approaches from the literature.

15: \end{abstract}

16: