A similarity measure for complex numbers

Research output: Unpublished contribution to conferencePoster

Abstract

Complex numbers are encountered in many applications due to integral
transforms
such as Fourier or Laplace or simply due to implicit wave phenomena. But
while
countless similarity measures exist for real valued data, its difficult to
compare two complex datasets. One has to resort to either comparing the data
in
parts, e.g. real/imaginary or absolute value/phase, or resort to error
measures
based on the mean squared error (MSE). Often applications provide a good
intuition of which mismatches are tolerable. We therefore propose a versatile
similarity measure that allows an easy local and global interpretation. It
is specifically crafted for complex-valued data and guarantees scale
invariance, invariance under constant phase shifts, symmetry with respect to
the arguments, but can be freely adjusted to match the desired notion of
similarity. We show that it globally matches the MSE and hope to provide a
useful template for anybody required to compare complex-valued datasets.
Original languageEnglish
Publication statusPublished - 8 Jun 2017
EventNSE PhD Day 2017 - Brussels, Belgium
Duration: 8 Jun 2017 → …

Conference

ConferenceNSE PhD Day 2017
CountryBelgium
CityBrussels
Period8/06/17 → …

Fingerprint

Dive into the research topics of 'A similarity measure for complex numbers'. Together they form a unique fingerprint.

Cite this