Value Driven Information Processing and Fusion
Abstract
The objective of the project is to develop a general framework for value driven decentralized information processing. Instead of attempting to identify a unifying information metric for network inference, our approach is to develop a framework that is applicable to various information value metrics as called for by different inference tasks. Major theoretical breakthroughs have been obtained under this effort, including: optimal data reduction in a network setting for decentralized inference with quantization constraint; interactive fusion that allows queries and interactive information exchange in either tandem or parallel networks; new operational interpretation of Wyners common information when information loss is inevitable; quantizer design for decentralized estimation; distributed network consensus and multi-agent optimization. The project has enriched the literature in information driven decentralized inference; more importantly, new challenges for inference over networks have been identified that may have broad ramifications in various emerging big data settings when inference is often hampered by practical constraints on information exchange.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2016
- Accession Number
- AD1009840
Entities
People
- Biao Chen
Organizations
- Syracuse University