SigSag: Iterative Detection Through Soft Message-Passing

Minimizing Message Passing in Parallel Iterative Codes

A compiler transformation to minimize the number of messages in PDE parallel iterative codes is presented. For the n-dimensional high-order PDE or second-order PDE using high-order finite differences, this transformation reduces the number of messages from as large as 3 − 1 to the minimum 2n.

Message Passing for Soft Constraint Dual Decomposition

Dual decomposition provides the opportunity to build complex, yet tractable, structured prediction models using linear constraints to link together submodels that have available MAP inference routines. However, since some constraints might not hold on every single example, such models can often be improved by relaxing the requirement that these constraints always hold, and instead replacing the...

A Modified Nonparametric Message Passing Algorithm for Soft Iterative Channel Estimation

Based on the factor graph framework, we derived a Modified Nonparametric Message Passing Algorithm (MNMPA) for soft iterative channel estimation in a Low Density Parity-Check (LDPC) coded Bit-Interleaved Coded Modulation (BICM) system. The algorithm combines ideas from Particle Filtering (PF) with popular factor graph techniques. A Markov Chain Monte Carlo (MCMC) move step is added after typica...

Neighborhood beautification Graph layout through message passing

Graph layout algorithms are used to compute aesthetic and useful visualizations of graphs. In general, for graphs with up to a few hundred nodes, force-directed layout algorithms produce good layouts. Unfortunately, for larger graphs, they often get stuck at local minima and have high computational complexity. In this paper, we introduce a novel message passing technique for graph layout. The k...

Failure Detection in Message Passing Systems