Page - 443 - in Differential Geometrical Theory of Statistics
Image of the Page - 443 -
Text of the Page - 443 -
Entropy2016,18, 110
2.2. ParameterSpoiling
In the theoryoferror-correctingcodes,oneconsiders spoilingoperationsonthecodeparameters.
Appliedtoan [n,k,d]2-codeC, theseproduce, respectively,newcodeswith the followingdescription
(seeSection1.1.1of [24]):
⢠A code C1 = C i f in Fn+12 , for a map f : C â F2, whose code words are of the form
(x1, . . . ,xiâ1, f(x1, . . . ,xn),xi, . . . ,xn) forw = (x1, . . . ,xn) â C. If f is a constant function, C1
isan [n+1,k,d]2-code. If allpairsw,wⲠâCwithdH(w,wâ˛)= dhave f(w) = f(wâ˛), thenC1 is
an [n+1,k,d+1]2-code.
⢠AcodeC2=C i inFnâ12 ,whosecodewordsaregivenbytheprojections
(x1, . . . ,xiâ1,xi+1, . . . ,xn)
ofcodewords (x1, . . . ,xiâ1,xi,xi+1, . . . ,xn) inC. This isan [nâ1,k,dâ1]2-code,exceptwhenall
pairsw,wⲠâCwithdH(w,wâ˛)= dhavethesameletterxi, inwhichcase it isan [nâ1,k,d]2-code.
⢠AcodeC3 = C(a, i)â C â Fn2, givenby the level setC(a, i) = {w = (xk)nk=1 â C | xi = a}.
TakingC(a, i) igivesan [nâ1,kâ˛,dâ˛]2-codewithkâ1⤠kâ˛< k, anddⲠ⼠d.
Thesamespoilingoperationsholdforq-arycodesCâFnq, foranyďŹxedq.
Inoursetting,whereC is thecodeobtainedfromafamilyof languages,accordingtotheprocedure
describedabove, theďŹrst spoilingoperationcanbeseenas theeffectof consideringonemoresyntactic
parameter,which isdependenton theotherparameters, hencedescribinga function F :Fn2 âF2,
whose restriction toC gives the function f : C â F2. In particular, the casewhere f is constant
onC represents the situation inwhich thenewparameter addsnouseful comparison information
for theselectedfamilyof languages. Thesecondspoilingoperationconsists in forgettingoneof the
parameters, and the thirdcorresponds to formingsubfamiliesof thegiven familyof languages, by
groupingtogether those languageswithasetvalueofoneof thesyntacticparameters. Thus,all these
spoilingoperationshaveaclearmeaningfromthepointofviewof the linguisticPCM.
2.3. Examples
Weconsider the same list of 63 parameters used in [3] (see Section 5.3.1 and TableA). This
choiceofparameters followsthemodularizedglobalparameterizationmethodof [2], for theDeterminer
Phrasemodule. They encompass parameters dealingwith person, number, and gender (1â6 on
their list), parameters of deďŹniteness (7â16 in their list), of countability (17â24), genitive structure
(25â31), adjectivalandrelativemodiďŹcation (32â14),positionandmovementof theheadnoun(42â50),
demonstrativesandotherdeterminers(51â50and60â63),possessivepronouns(56â59); seeSection5.3.1
andSection5.3.2of [3] formoredetails.
Our very simple examples here are justmeant to clarify our notation: they consist of some
collectionsof languagesselectedfromthelistof28,mostlyIndo-European, languagesconsideredin[3].
Ineachgroupweconsiderweeliminate theparameters thatareentailed fromothers, andwefocuson
ashorter list, amongtheremainingparameters, thatwill sufďŹce to illustrateourviewpoint.
Example1. ConsideracodeC formedoutofthelanguages 1= Italian, 2=Spanish,and 3=French,
andletusconsideronly theďŹrst sixsyntacticparametersofTableAof [3], so thatCâFn2 withn=6.
Thecodewords for the three languagesare
1 1 1 1 0 1 1
2 1 1 1 1 1 1
3 1 1 1 0 1 0
Thishascodeparameters (R= log2(3)/6=0.2642,δ=1/6),whichsatisfyR<1âH2(δ),hence
they liebelowtheGVcurve (seeEquation(8)below).Weuse thiscodeto illustrate the threespoiling
operationsmentionedabove.
443
Differential Geometrical Theory of Statistics
- Title
- Differential Geometrical Theory of Statistics
- Authors
- FrĂŠdĂŠric Barbaresco
- Frank Nielsen
- Editor
- MDPI
- Location
- Basel
- Date
- 2017
- Language
- English
- License
- CC BY-NC-ND 4.0
- ISBN
- 978-3-03842-425-3
- Size
- 17.0 x 24.4 cm
- Pages
- 476
- Keywords
- Entropy, Coding Theory, Maximum entropy, Information geometry, Computational Information Geometry, Hessian Geometry, Divergence Geometry, Information topology, Cohomology, Shape Space, Statistical physics, Thermodynamics
- Categories
- Naturwissenschaften Physik