English

A Class of Optimal Structures for Node Computations in Message Passing Algorithms

Information Theory 2021-10-12 v3 Hardware Architecture math.IT

Abstract

Consider the computations at a node in a message passing algorithm. Assume that the node has incoming and outgoing messages x=(x1,x2,,xn)\mathbf{x} = (x_1, x_2, \ldots, x_n) and y=(y1,y2,,yn)\mathbf{y} = (y_1, y_2, \ldots, y_n), respectively. In this paper, we investigate a class of structures that can be adopted by the node for computing y\mathbf{y} from x\mathbf{x}, where each yj,j=1,2,,ny_j, j = 1, 2, \ldots, n is computed via a binary tree with leaves x\mathbf{x} excluding xjx_j. We make three main contributions regarding this class of structures. First, we prove that the minimum complexity of such a structure is 3n63n - 6, and if a structure has such complexity, its minimum latency is δ+log(n2δ)\delta + \lceil \log(n-2^{\delta}) \rceil with δ=log(n/2)\delta = \lfloor \log(n/2) \rfloor, where the logarithm always takes base two. Second, we prove that the minimum latency of such a structure is log(n1)\lceil \log(n-1) \rceil, and if a structure has such latency, its minimum complexity is nlog(n1)n \log(n-1) when n1n-1 is a power of two. Third, given (n,τ)(n, \tau) with τlog(n1)\tau \geq \lceil \log(n-1) \rceil, we propose a construction for a structure which we conjecture to have the minimum complexity among structures with latencies at most τ\tau. Our construction method runs in O(n3log2(n))O(n^3 \log^2(n)) time, and the obtained structure has complexity at most (generally much smaller than) nlog(n)2n \lceil \log(n) \rceil - 2.

Keywords

Cite

@article{arxiv.2009.02535,
  title  = {A Class of Optimal Structures for Node Computations in Message Passing Algorithms},
  author = {Xuan He and Kui Cai and Liang Zhou},
  journal= {arXiv preprint arXiv:2009.02535},
  year   = {2021}
}

Comments

Accepted for publication in IEEE Transactions on Information Theory

R2 v1 2026-06-23T18:20:04.453Z