edmcr: Euclidean Distance Matrix Completion Tools

Implements various general algorithms to estimate missing elements of a Euclidean (squared) distance matrix. Includes optimization methods based on semi-definite programming found in Alfakih, Khadani, and Wolkowicz (1999)<doi:10.1023/A:1008655427845>, a non-convex position formulation by Fang and O'Leary (2012)<doi:10.1080/10556788.2011.643888>, and a dissimilarity parameterization formulation by Trosset (2000)<doi:10.1023/A:1008722907820>. When the only non-missing distances are those on the minimal spanning tree, the guided random search algorithm will complete the matrix while preserving the minimal spanning tree following Rahman and Oldford (2018)<doi:10.1137/16M1092350>. Point configurations in specified dimensions can be determined from the completions. Special problems such as the sensor localization problem, as for example in Krislock and Wolkowicz (2010)<doi:10.1137/090759392>, as well as reconstructing the geometry of a molecular structure, as for example in Hendrickson (1995)<doi:10.1137/0805040>, can also be solved. These and other methods are described in the thesis of Adam Rahman(2018)<https://hdl.handle.net/10012/13365>.

Version: 0.2.0
Depends: R (≥ 3.2.0)
Imports: Matrix, igraph, lbfgs, truncnorm, MASS, nloptr, vegan, sdpt3r, utils, methods, stats
Published: 2021-09-10
Author: Adam Rahman [aut], R. Wayne Oldford [aut, cre, ths]
Maintainer: R. Wayne Oldford <rwoldford at uwaterloo.ca>
License: GPL-2 | GPL-3
URL: https://github.com/great-northern-diver/edmcr
NeedsCompilation: yes
Materials: README NEWS
CRAN checks: edmcr results


Reference manual: edmcr.pdf


Package source: edmcr_0.2.0.tar.gz
Windows binaries: r-devel: edmcr_0.2.0.zip, r-devel-UCRT: edmcr_0.2.0.zip, r-release: edmcr_0.2.0.zip, r-oldrel: edmcr_0.2.0.zip
macOS binaries: r-release (arm64): edmcr_0.2.0.tgz, r-release (x86_64): edmcr_0.2.0.tgz, r-oldrel: edmcr_0.2.0.tgz
Old sources: edmcr archive


Please use the canonical form https://CRAN.R-project.org/package=edmcr to link to this page.