1: \begin{abstract}

2: We derive and study SQMC (Sequential Quasi-Monte Carlo), a class of algorithms

3: obtained by introducing QMC point sets in particle filtering. 

4: SQMC is related to, and may be seen as an extension of, the array-RQMC algorithm of \cite{LEcuyer2006}. 

5: The complexity of SQMC is $\bigO(N\log N)$, where $N$ is

6: the number of simulations at each iteration, and its error rate is

7: smaller than the Monte Carlo rate $\bigO_P(N^{-1/2})$. The only requirement

8: to implement SQMC is the ability to write the simulation of particle

9: $\bx_{t}^{n}$ given $\bx_{t-1}^{n}$ as a deterministic function

10: of $\bx_{t-1}^{n}$ and a fixed number of uniform variates. 

11: We show that SQMC is amenable

12: to the same extensions as standard SMC, such as forward smoothing,

13: backward smoothing, unbiased likelihood evaluation, and so on. In

14: particular, SQMC may replace SMC within a PMCMC (particle Markov chain

15: Monte Carlo) algorithm. We establish several convergence results.

16: We provide numerical evidence that 

17: SQMC may significantly outperform SMC in practical scenarios.

18: 

19: Key-words: Array-RQMC; Low discrepancy; Particle filtering; Quasi-Monte Carlo;

20: Randomized Quasi-Monte Carlo; Sequential Monte Carlo

21: \end{abstract}

22: