1: \begin{abstract}

2:   \citet{SQMC} recently introduced Sequential quasi-Monte Carlo (SQMC)

3:   algorithms as an efficient way to perform filtering in state-space

4:   models.  The basic idea is to replace random variables with

5:   low-discrepancy point sets, so as to obtain faster convergence

6:   than with standard particle filtering.  \citet{SQMC} describe briefly

7:   several ways to extend SQMC to smoothing, but do not provide supporting

8:   theory for this extension.  We discuss more

9:   thoroughly how smoothing may be performed within SQMC, and derive

10:   convergence results for the so-obtained smoothing algorithms. We

11:   consider in particular SQMC equivalents of  forward smoothing and

12:   forward filtering backward sampling, which are the most well-known smoothing techniques.

13:   As a preliminary step, we provide a generalization of the classical result of \citet{Hlawka1972} on the

14:   transformation of QMC point sets into low discrepancy point sets

15:   with respect to non uniform distributions. As a corollary of the

16:   latter, we note that we can slightly weaken the assumptions to prove

17:   the consistency of SQMC.

18: 

19:   \textit{Keywords}: Hidden Markov models; Low discrepancy; Particle

20:   filtering; Quasi-Monte Carlo; Sequential quasi-Monte Carlo;

21:   Smoothing; State-space models.

22: 

23: \end{abstract}

24: