1: \begin{abstract}

2: Low density lattice codes (LDLC) are a family of lattice codes that can be

3: decoded efficiently using a message-passing algorithm. In the original LDLC

4: decoder, the message exchanged between variable nodes and check nodes are

5: continuous functions, which must be approximated in practice. A promising

6: method is Gaussian approximation (GA), where the messages are approximated

7: by Gaussian functions. However, current GA-based decoders share two

8: weaknesses: firstly, the convergence of these approximate decoders is

9: unproven; secondly, the best known decoder requires $O(2^d)$ operations at

10: each variable node, where $d$ is the degree of LDLC. It means that existing

11: decoders are very slow for long codes with large $d$. The contribution of this

12: paper is twofold: firstly, we prove that all GA-based LDLC decoders converge

13: sublinearly (or faster) in the high signal-to-noise ratio (SNR) region;

14: secondly, we propose a novel GA-based LDLC decoder which requires only $O(d)$

15: operations at each variable node. Simulation results confirm that

16: the error correcting performance of proposed decoder is the same

17: as the best known decoder, but with a much lower decoding complexity.

18: 

19: % Given the same amount of computation resource, the proposed decoder can afford much more iterations that the best known one, thus provide better decoding performance. Simulation results confirm that the proposed decoder with $20$ iterations outperforms the best known one with $10$ iterations.

20: %the same as that of the best known precoders.

21: 

22: %and approach the capacity of the additive white Gaussian noise (AWGN) channel.

23: %but with a much lower decoding complexity. Simulation

24: %results confirm that the error performance of XY-precoder is almost

25: %the same as that of the best known precoders.

26: \end{abstract}

27: