Seite - 312 - in Differential Geometrical Theory of Statistics
Bild der Seite - 312 -
Text der Seite - 312 -
entropy
Article
ASequenceofEscortDistributionsand
GeneralizationsofExpectationson
q-ExponentialFamily
HiroshiMatsuzoe
DepartmentofComputerScienceandEngineering,NagoyaInstituteofTechnology,Nagoya466-8555, Japan;
matsuzoel@nitech.ac.jp;Tel.: +81-52-735-5143
AcademicEditors: FrédéricBarbarescoandFrankNielsen
Received: 26October2016;Accepted: 19December2016;Published: 25December2016
Abstract: In the theoryofcomplexsystems, longtailedprobabilitydistributionsareoftendiscussed.
Forsuchaprobabilitydistribution,adeformedexpectationwithrespect toanescortdistribution is
moreuseful thanthestandardexpectation. In thispaper,bygeneralizingsuchescortdistributions,
a sequence of escort distributions is introduced. As a consequence, it is shown that deformed
expectationswithrespect tosequentialescortdistributionseffectivelyworkforanomalousstatistics.
In particular, it is shown that a Fisher metric on a q-exponential family can be obtained from
the escort expectation with respect to the second escort distribution, and a cubic form (or an
Amari–Chentsov tensor field, equivalently) is obtained from the escort expectationwith respect
to the thirdescortdistribution.
Keywords: escortdistribution;escortexpectation; statisticalmanifold;deformedexponential family;
Tsallis statistics; informationgeometry
MSC:53A15;53B50;62F99;94A14
1. Introduction
Long tailedprobabilitydistributionsand their relatedprobabilitydistributionsare important
objects in anomalous statistical physics (cf. [1–3]). For such long tailed probability distributions,
the standardexpectationdoesnot exist ingeneral. Therefore, thenotionof escortdistributionhas
been introduced[4]. Sinceanescortdistributiongivesasuitableweight for tailprobability, theescort
expectationwhich is the expectationwith respect to anescortdistribution ismoreuseful than the
standardone.
Inanomalousstatistics, adeformedexponential functionandadeformedlogarithmfunctionplay
essential roles. In fact, adeformedexponential family isan importantstatisticalmodel inanomalous
statistics. Suchastatisticalmodel isdescribedbysuchadeformedexponential function. Inparticular,
thesetofallq-normaldistributions(orStudent’s t-distributions,equivalently) isaq-exponential family,
which isdescribedbyaq-deformedexponential function[5] (seealso [6,7]).
Ontheotherhand,ageneralizedscore function isdefinedfromadeformedlogarithmfunction.
In thepreviousworks, theauthorshowedthatadeformedscore function isunbiasedwithrespect to
theescortexpectation[8,9]. This implies thatadeformedscore function is regardedasanestimating
functiononadeformedexponential family. Inaddition, in informationgeometry, it is knownthat
adeformedexponential familyhasastatisticalmanifoldstructure. Thenadeformedscore function
is regarded as a tangent vector on this statisticalmanifold [6,10]. Therefore, properties of escort
expectationsarecloselyrelatedtogeometric structuresonadeformedexponential family.
Entropy2017,19, 7 312 www.mdpi.com/journal/entropy
Differential Geometrical Theory of Statistics
- Titel
- Differential Geometrical Theory of Statistics
- Autoren
- Frédéric Barbaresco
- Frank Nielsen
- Herausgeber
- MDPI
- Ort
- Basel
- Datum
- 2017
- Sprache
- englisch
- Lizenz
- CC BY-NC-ND 4.0
- ISBN
- 978-3-03842-425-3
- Abmessungen
- 17.0 x 24.4 cm
- Seiten
- 476
- Schlagwörter
- Entropy, Coding Theory, Maximum entropy, Information geometry, Computational Information Geometry, Hessian Geometry, Divergence Geometry, Information topology, Cohomology, Shape Space, Statistical physics, Thermodynamics
- Kategorien
- Naturwissenschaften Physik