10000 simulations from a stochastic Lotka-Volterra model, assigned weights according to a Gaussian approximate Bayesian computation kernel with tolerance equal to 50.
The output of an SMC sampler where the initial distribution is a Gaussian and the final target is a mixture of Gaussians. 25 particles were used, with an adaptive method to determine the sequence of targets, and a Metropolis-Hastings move to move the particles at each step.