Accelerating sequential Monte Carlo with surrogate likelihoods

Delayed-acceptance is a technique for reducing computational effort Bayesian models with expensive likelihoods. Using delayed-acceptance kernel Markov chain Monte Carlo can reduce the number of likelihoods evaluations required to approximate posterior expectation. uses surrogate, or approximate, likelihood avoid evaluation when possible. Within sequential framework, we utilise history sampler adaptively tune surrogate yield better approximations and use first annealing schedule further increase efficiency. Moreover, propose framework optimising computation time whilst avoiding particle degeneracy, which encapsulates existing strategies in literature. Overall, develop novel algorithm computationally efficient SMC functions. The method applied static models, demonstrate on toy real examples.