### Citations

1350 |
Information Theory: Coding Theorems for Discrete Memoryless Systems, Akadémiai Kiado
- Cśiszar, Körner
(Show Context)
Citation Context ...n1}. Then, (8) is equivalent to Ĥ(x̂n1 , x̂ n 2 ) ∈ T (R, δn). If there is more than one pair or no such pair in B1(m1) × B2(m2), declare a decoding error. Note that our decoding scheme is universal =-=[13]-=-, i.e., the decoder does not depend on knowledge of the true distribution pX1,X2 . Analysis of error probability: Let the sequences sent by the two users be (Xn1 , X n 2 ) and let their corresponding ... |

1262 |
Noiseless coding of correlated information sources
- Slepian, Wolf
- 1973
(Show Context)
Citation Context ...,M2). Subsequently, given these compressed versions of the sources, a decoder seeks to reconstruct (Xn1 , X n 2 ). One of the most remarkable results in information theory, proved by Slepian and Wolf =-=[1]-=-, states that the set of achievable rate pairs (R1, R2) is equal to that when each of the encoders is given knowledge of the other source, i.e., encoder 1 knows Xn2 and vice versa. The optimal rate re... |

161 |
Information-Spectrum Methods in Information Theory.
- Han
- 2002
(Show Context)
Citation Context ...t part of Theorem 1 is proved using the usual random binning argument together with a multidimensional Berry-Essèen theorem [2]. The converse is proved using an information spectrum technique by Han =-=[12]-=-. Theorem 1 extends to the case where there are more than two senders. By examining Rin(n, ) and Rout(n, ), it can be seen that we have characterized the (n, )-rate region up to an O( lognn ) facto... |

44 | Fixed-length lossy compression in the finite blocklength regime,” - Kostina, Verdu - 2012 |

42 |
On the dependence of the Berry–Esseen bound on dimension.
- Bentkus
- 2003
(Show Context)
Citation Context ...ernational B.V. the finite blocklength, fixed error probability regime, which is clearly of relevance in practice. The main tool that we use is a multidimensional version of the Berry-Essèen theorem =-=[2]-=-. A. Main Contributions This paper characterizes the (n, )-optimal rate region for the SW problem R∗(n, ) up to an O( lognn ) factor. In the course of doing so, we introduce a fundamental quantity c... |

37 |
Second-order asymptotics in fixed-length source coding and intrinsic randomness
- Hayashi
- 2008
(Show Context)
Citation Context ...sources are symmetric. This work generalizes their setting in that we consider all discrete sources. This paper is a network information theory analogue of the works on second-order coding rates [6], =-=[7]-=- and finite blocklength analysis [8]–[11]. II. PROBLEM STATEMENT AND MAIN RESULTS A. Notation Random variables and the values they take on will be denoted by upper case (e.g., X) and lower case (e.g.,... |

20 | The dispersion of lossy source coding,” - Ingber, Kochman - 2011 |

20 | The dispersion of joint sourcechannel coding.
- Wang, Ingber, et al.
- 2011
(Show Context)
Citation Context ...izes their setting in that we consider all discrete sources. This paper is a network information theory analogue of the works on second-order coding rates [6], [7] and finite blocklength analysis [8]–=-=[11]-=-. II. PROBLEM STATEMENT AND MAIN RESULTS A. Notation Random variables and the values they take on will be denoted by upper case (e.g., X) and lower case (e.g., x) respectively. Types (empirical distri... |

16 | On the redundancy of Slepian-Wolf coding
- He, Lastras-Montaño, et al.
- 2009
(Show Context)
Citation Context ...re not included this paper but will be discussed in an extended version of the current work. B. Related Work The redundancy of SW coding was discussed by Baron et al. [3] and followed up by He et al. =-=[4]-=-. However, in these works, the authors considered a single source X1 to be compressed and side information X2 available only at the decoder. Thus, X2 is neither coded nor estimated. They showed that a... |

13 |
Channel coding in the finite blocklength regime
- Polyanskiy, Poor, et al.
- 2010
(Show Context)
Citation Context ...eralizes their setting in that we consider all discrete sources. This paper is a network information theory analogue of the works on second-order coding rates [6], [7] and finite blocklength analysis =-=[8]-=-–[11]. II. PROBLEM STATEMENT AND MAIN RESULTS A. Notation Random variables and the values they take on will be denoted by upper case (e.g., X) and lower case (e.g., x) respectively. Types (empirical d... |

12 |
Asymptotische Abschätzungen
- Strassen
- 1962
(Show Context)
Citation Context ... the sources are symmetric. This work generalizes their setting in that we consider all discrete sources. This paper is a network information theory analogue of the works on second-order coding rates =-=[6]-=-, [7] and finite blocklength analysis [8]–[11]. II. PROBLEM STATEMENT AND MAIN RESULTS A. Notation Random variables and the values they take on will be denoted by upper case (e.g., X) and lower case (... |

11 | Redundancy rates of Slepian-Wolf coding
- Baron, Khojastepour, et al.
- 2004
(Show Context)
Citation Context ...y mass of the degenerate Gaussian N (0,V) lies in a subspace of dimension one. Therefore, the set S (V, ) = {z ∈ R3 : z ≥ √vQ−1()1} is axis-aligned. The quantity √ v nQ −1() is the rate redundancy =-=[3]-=-–[5] for fixed-length SW coding in the finite blocklength regime for a DMMS for which rank(V) = 1. In this case, the bounds in (5) and (6) (up to O( lognn ) factors) degenerates to R ≥ H + √ v n Q−1(... |

10 | Inequalities for the L1 deviation of the empirical distribution. Hewlett-Packard Labs, - Weissman, Ordentlich, et al. - 2003 |

6 | Non-asymptotic performance of symmetric Slepian-Wolf coding
- Sarvotham, Baron, et al.
- 2005
(Show Context)
Citation Context ...nalyzed a variable-length variant of the SW problem and showed that the dispersion is smaller than in the fixed-length setting. This dispersion is similar to that for channel coding. Sarvotham et al. =-=[5]-=- considered the SW problem with two sources to be compressed but limited their setting to the case the sources are symmetric. This work generalizes their setting in that we consider all discrete sourc... |