We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Information Theory

Title: Non-parametric Message Important Measure: Storage Code Design and Transmission Planning for Big Data

Abstract: Storage and transmission in big data are discussed in this paper, where message importance is taken into account. Similar to Shannon Entropy and Renyi Entropy, we define non-parametric message important measure (NMIM) as a measure for the message importance in the scenario of big data, which can characterize the uncertainty of random events. It is proved that the proposed NMIM can sufficiently describe two key characters of big data: rare events finding and large diversities of events. Based on NMIM, we first propose an effective compressed encoding mode for data storage, and then discuss the channel transmission over some typical channel models. Numerical simulation results show that using our proposed strategy occupies less storage space without losing too much message importance, and there are growth region and saturation region for the maximum transmission, which contributes to designing of better practical communication system.
Comments: 30 pages one-colunm, 9 figures
Subjects: Information Theory (cs.IT); Statistics Theory (math.ST)
Cite as: arXiv:1709.10280 [cs.IT]
  (or arXiv:1709.10280v1 [cs.IT] for this version)

Submission history

From: Shanyun Liu [view email]
[v1] Fri, 29 Sep 2017 08:22:31 GMT (919kb)

Link back to: arXiv, form interface, contact.