1: \begin{abstract}

2:     We propose a novel method for sampling and optimization tasks based on a stochastic interacting particle system. We explain how this method can be used for the following two goals:

3:     (i) generating approximate samples from a given target distribution;

4:     (ii) optimizing a given objective function. The approach is derivative-free and affine invariant, and is therefore well-suited for solving inverse problems defined by complex

5:     forward models: (i) allows generation of samples from the Bayesian posterior and (ii) allows determination of the maximum a posteriori estimator.

6:     % allowing (i) to understand the distribution of parameters taking into account both the underlying model and the available data; or (ii) to estimate the most likely parameter given the available data and the model relating them.

7:     We investigate the properties of the proposed family of methods in terms of various parameter choices, both analytically and by means of numerical simulations. The analysis and numerical

8:     simulation establish that the method has potential for

9:     general purpose optimization tasks over Euclidean space;

10:     contraction properties of the algorithm are established under suitable conditions, and computational experiments demonstrate

11:     wide basins of attraction for various specific problems.

12:     The analysis and experiments

13:     also demonstrate the potential for the sampling methodology

14:     in regimes in which the target distribution is unimodal

15:     and close to Gaussian; indeed we prove that the method recovers

16:     a Laplace approximation to the measure in certain

17:     parametric regimes and provide numerical evidence that this

18:     Laplace approximation attracts a large set of initial conditions in a number of examples.

19:     %In particular, we prove convergence of the method in simple %settings, Gaussian or one-dimensional.

20: \end{abstract}

21: