Description
TL;DR: A mathematical treatment of distributed averaging when messages are quantized, links fail randomly, and finite quantizers can saturate.
About This Document
Soummya Kar and José M. F. Moura add dither before quantization and analyze the resulting sensor-network consensus process through stochastic approximation. For an unbounded quantizer they establish almost-sure and mean-square convergence to a finite random value, with an accuracy-versus-speed tradeoff controlled by link weights. For bounded quantizers they derive probability bounds and design choices involving level count, step size, saturation risk, and target accuracy.
Why Bitcoin People May Care
This is not a Bitcoin paper and its nodes are not miners choosing an adversarial transaction history. It supplies pre-Bitcoin distributed-systems background: even honest participants can struggle to converge when communication is lossy and information is compressed. The paper is useful for comparing failure models and definitions of consensus without pretending that average consensus and Nakamoto consensus solve the same problem.
What You Receive
Purchase reveals the verified preserved PDF location for this document.
Document Details
- Title: Distributed Consensus with Quantized Data and Random Link Failures
- Author / organization: Soummya Kar and José M. F. Moura
- Year: 2009
- Language: English
- Document type: Distributed-systems research paper
- Pages: 54
CacheRat provides researched source-location and document-identification information. CacheRat does not claim ownership of the underlying third-party documentation.
