The wiring economy principle for designing inference networks

Research output: Contribution to journalArticlepeer-review

Abstract

The wiring economy principle in neuroscience has explained many experimentally observed properties of neuronal networks by asserting the need to keep the axons and dendrites that connect neurons small in length. Just like neuronal networks, many distributed systems are physical constructs that incur deployment and maintenance costs for their communication infrastructure. Taking wiring economy as a design goal for engineering systems that perform distributed coordination and inference, this paper formulates and studies the tradeoff between performance and wiring cost. It is shown that separated communication topology design and physical node placement yields optimal design. Designing optimal networks is shown to be NP-complete. The natural relaxation to the integer network design problem is shown to be a reverse convex program. Small optimal networks are computed. Optimally placed random network topologies are demonstrated to have good performance.

Original languageEnglish (US)
Article number6517113
Pages (from-to)1095-1104
Number of pages10
JournalIEEE Journal on Selected Areas in Communications
Volume31
Issue number6
DOIs
StatePublished - May 29 2013
Externally publishedYes

Keywords

  • Distributed inference
  • network design
  • wiring

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'The wiring economy principle for designing inference networks'. Together they form a unique fingerprint.

Cite this