Sei sulla pagina 1di 8

22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

Casa TiendadeVideoTraining&Libros Radar SafariBooksOnline Conferencias CursosdeTIyCertificados

Datos Mstemas

Paralelismossorprendentesentrelas
matemticasylaingenierade
software
Cadavezmsfamiliarizadosconlasmatemticasleayudar
cruzpolinizanideasentrelasmatemticasylaingenieradesoftware.
porAliceZheng|Comentarios:25|19deenero2015

Notadeleditor:AliceZhengserpartedelequipodeenseanzaagranescalaMquinaDa
deAprendizajeenStrata+HadoopMundialenSanJos.VisiteelsitiowebStrata+
HadoopMundialparaobtenermsinformacinsobreelprograma.

Durantemiprimeraoenlauniversidad,tuveunaepifanasobrelasmatemticasque
cambiarontodamiperspectivasobreelcampo.Habaelegidoparaestudiarelaprendizaje
demquina,unreadeinvestigacininterdisciplinarquecombinaelementosdela
informtica,estadstica,ynumerosossubcamposdelasmatemticas,comola
optimizacinyellgebralineal.Fueunmuchoparatomar,ytodosnosotroslosestudiantes
deprimeraoestabaluchandoparaabsorberlaavalanchadenuevosconceptos.

Unanoche,estabasentadoenlaoficinatratandodeasimilarellgebralineal.Unlibrode
textomaravillosamentelcidomesirvidegua:IntroduccinallgebraLineal,escritopor
GilbertStrang.Peroyonoestabahaciendol.Estababuscandoendiversasdefiniciones
descomposicineigen,Jordaniaformascannicas,lainversindematrices,etc.ypens:
"Porqu?"Porqutodoparecetanraro?Porqueselinversodefinedeestamanera?
Ahoraquelopienso,porqusoncualquieradelamatrizdelasoperacionesdefinidas
comoestn?

Aunquemirandoaunaparedsinesperanzadesmbolos,unrayodeluzseencendienmi
mente.Yotenaunaidea:lasmatemticassonundiseo.Antesdeesemomento,me
habaacercadoalasmatemticascomosifueraverdaduniversal:trascendenteensu
perfeccin,casiimposibledeconocerporlossimplesmortales.Peroenesanoche,medi
cuentadequelasmatemticassonunaherramientahumanoconstruido.Mathest
diseado,aligualquelosprogramasdesoftwareestndiseados,yelusodemuchosde
losmismosprincipiosdediseo.Estosprincipiospuedennoserevidentes,peroson
comprensibles.Enesemomento,lasmatemticaspasarondeserincognosciblepara
razonable.

Lasmatemticassonunsistemadeobjetos,operacionesyrepresentacionesdela
taquigrafa.Estdiseadoparamodelarfenmenosdelmundoreal.Comotodoslos
diseos,hayciertosgradosdelibertad.Elsistemapodrahabersidoconstruidodeuna
manerauotra.Unamatrizpodrahabersidodiseadocomounapelotaredonda,en
coordenadaspolares.Noimporta,siempreycuandolasoperacionessonconsistenteses
slounaabreviatura.Enalgnmomento,alguienhizoesasdecisionesdediseo.
Recogieronlosobjetosylasoperaciones,yestablecennormasdeorganizacin.Sobrela

http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 1/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

basedeestasdecisionesfundamentalessisediseanbienunnmerodeotras,
propiedadesdemostrablestilesdespusseguir,ytodalacosasepuedeutilizarpara
modelarlascosasqueexperimentamosenelmundoreal:laformaenqueunapelota
lanzadaviajaatravsdeespacio,lasondasdesonidoatravsdelaformaenquese
lanzanter,elascensoylacadadepreciosdelasacciones.Larealidadfsicacontiene
capasycapasdecomplejidad.Sistemasmatemticosbiendiseadosofrecen
herramientaslimpiasyconcisaspararepresentarlarealidadfsicaencadacapa.

lgebralinealestdiseadopararepresentaralossistemasdeecuacioneslineales.
Ecuacioneslinealesestndiseadospararepresentarrelacioneslineales,cuandouna
entidadestescritoparaserunasumademltiplosdeotrasentidades.Enlataquigrafade
lgebralineal,unarelacinlinealserepresentacomounoperadorlinealunamatriz.
Operadoreslinealessehacenparasersimple,porloquesusefectospuedenser
completamenteanalizados.Sepuedenhacerdoscosas:rotacinylaescala.Aqu,enla
fronteradelgebraygeometra,unpocodelamagiasucede.Lasoperacionesalgebraicas
demultiplicacinysumasetraducenenrotacinylaescaladelosvectoresenelespacio
vectorial.Estonospermiteanalizarlosefectosgeomtricosdeunoperadorlinealutilizando
ellgebra,rompiendolamatrizensuspartesconstituyentes:lacantidadderotacin,la
cantidaddeestiramientoocompresin,yenelquelasdirecciones.Laconstruccinde
lgebralinealsirvecomounejemplodeunpatrnqueescomnentodalamatemtica:
cabodeunospocosobjetosyunconjuntodeoperacioneslimitadas,surgenpoderosas
propiedadesquepermitenunamejorcomprensindelaestructuradelproblemaascomo
solucioneseficientes.

Fuente:CortesadeAliceZheng

Eldiseodelasmatemticasabarcaunnmerodeotrosprincipiosquetambinestn
presentesenlaingenieradesoftware.Tomelgebraabstracta,porejemplo.Resumen
lgebraesesencialmenteunejerciciodediseodejerarquadeobjetos,dondeelobjetivo
esutilizarlamenorcantidaddeingredientescomoseaposible,aadiendouningrediente
msalavez,paraverqutipodeconstruccionesinteresantesytilesquepodamos
conseguir.Ungruposedefinecomounconjuntodeelementosjuntoconunaoperacin
(quetienequesatisfaceralgunascondicionesparaasegurarsedequesucomportamiento
noesdemasiadoextrao).Unanilloesuntipoespecialdegrupodotadodedos
operacionesquepuedenserconsideradoscomogeneralizacionesdeadiciny
multiplicacin.Uncampoesunanilloespecializadoconcuatrooperaciones
(generalizacionesdesuma,resta,multiplicacinydivisin).Estodebecomenzar
inmediatamenteasonarfamiliarauningenierodesoftware:esunajerarquadeobjetos,
dondeuncampoheredadeunanilloqueheredadeungrupo!

http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 2/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

Fuente:CortesadeAliceZheng

Echemosunvistazoalsistemadenmerosreales.Esteesotroejemplodediseode
objetosjerrquico,peroconalgunosgirosinteresantesenlosdetalles.Empezamosconlos
nmerosnaturales,locualesunaextensinnaturaldenuestrosdedos.Acontinuacin,
aadimossuimagenenelespejoatravsdelabrechacerolosnmerosnegativos.Esto
nosdalosnmerosenteros,loscuales,cuandosecombinanconlasoperacionesde
adicinymultiplicacin,formanunanillocomosehadefinidoanteriormente.Eltiroenlas
operacionesdemultiplicacinydivisin,yobtenemoslosnmerosracionales,quees
ahorauncampo.Lahistoriapodraterminaraqu,ytodoestarafeliz:tenemosunmontn
denmerosyunmontndeoperaciones,podemosaplicardichasoperacionesaesos
nmerosyanasterminarconelmismoconjuntodenmeros.Hurra!
Desafortunadamente,enestemomentonuestrovecinogemetrallamaanuestrapuertay
pregunta:"Qupasaconelreadeuncrculoolahipotenusadeuntringulorectngulo?
Esosnoparecehaberunarelacinentredosnmerosenteros".

ImagencortesadeAliceZheng

Estedescubrimientoabrelapuertadelainundacindelosnmerosirracionalesylanza
unallaveinglesaennuestrodiseolimpioyordenado.Tantolosracionalesylos
irracionalessonnmerostiles,yserabuenotenerunarepresentacinquepuedemanejar
ambos.Perosonesasbestiasdiferentes,tododebidoaunanocinmolestollamado
infinito:losirracionalessondelongitudinfinitayporlotantollegarasermuchomsdifcil
deprecisar.Losnmerosracionalessonordenadaycontable,mientrasquehay
uncountablytantasirracionales.Nostomunospocosmilesdeaosparaencontraruna
solucin.Ahoramismo,nuestramejorpropuestaesladerepresentartodocomounlmite:
http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 3/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

todonmerorealpuedeserpensadocomounaclasedeequivalenciadesecuenciasque
tienenelmismodestinoenelinfinito.Estaestructuraesjerrquica,ylamayoradelos
elementossondehechonocomputablesporunamquinadeTuringfinito.Alfinal,se
podradecirquenohaynadarealacercadelosnmerosrealesestodouna
construccin!

Estossonsloalgunosejemplosdediseomatemticoeneltrabajo.Nuestracultura
inculcalaideaextraadeque"lamatemticaesdifcil."Matemticasesvistocomo
demasiadoabstracto,demasiadoimpenetrable,demasiadodifcildedigerireimposiblede
saber.Perodesdeunaperspectivaalternativa,lasmatemticascontienesorprendentes
paralelismosconlaingenieradesoftware.Ambasdisciplinassonpesados enlajergayla

notacin.Perounavezqueseanalizanatravsdelajerga,podemosempezaraverla
carneyloshuesosdelasmatemticas.Lacomprensindelosprincipiosdediseodentro
delasmatemticasnosproporcionaunaentradaenestatierraextraadeobjetos
jerrquicosyelcambiodelasrepresentaciones.Alestarmsfamiliarizadoconelpaisaje
delasmatemticas,podemosayudarleconlapolinizacincruzadadeideasentrelas
matemticasylaingenieradesoftware.Talvezinclusopodemosempezarahacer
modificacionesyllegaranuevosdiseosdelasmatemticas.Hey,esesistemade
nmerosrealesseestbastanteviejoyengorroso.Listoparaalgonuevo?

ImagenrecortadaenartculosypginasdecategorasporTomBrownenFlickr,utilizada
bajounalicenciadeCreativeCommons.

Conexo

Elembalsededatos
MATLAB,R,yJulia:Idiomasparaelanlisisdedatos
Laautopsiadecortelibre
NoseradivertidoparaconstruirsupropiaGoogle?
ExplorarO'ReillyRadar

Java8interfacesfuncionales
Qulenguajedeberaustedaprenderprimero?
Equiposdecienciadedatosdeconstruccin
Elfuturodelaprogramacin
Impulsadopor
tags:BigData,cienciasdedatos,loscientficosdedatos,diseomatemtico,diseodesoftware,
ingenieradesoftware

RecibeelBoletndeDatosO'Reilly
Mantngaseinformado.Recibevisinsemanaldeexpertosdelaindustria.

Enteryouremail
Weprotectyourprivacy.

25Comments O'ReillyRadar Login

SortbyBest Share Favorite

Jointhediscussion

DavidWarman 3daysago
AthighschoolIwasfascinatedbypropositionallogicandpredicatecalculus.
PrivatestudyledmealsointoBooleanAlgebra.InparallelIwasreadingaboutthe
TheoryofFiniteAutomata(TrachtenbrotandKorzybski(?1967).Iwasalso
seriouslyintoelectronicshacking.Butthesewereseparatedendeavors.Iknew
OhmsLawandthesimpleANDlogicmadewithadiodeandtworesistors,butI
rememberittakingagoodyeartogrokhowthatrelatedtoBooleanAlgebra.Ido
notrememberwhatitwasliketonotknowbutinaninstant,aflash,I
comprehendedvisuallythereisa1to1mappingbothwaysbetweenthe
mathematicaltheoriesandtheflowofelectronsinthephysicalcircuits.

ThiswasandstillispureMagicforme.Iseenothinghappeningyetmuchisgoing
on.
http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 4/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

on.

WithinamonthIhaddesignedabitserialfouraccumulator7.7floatingpointALU.
Onpaper.Ittookabout900gates.Ofcourse,Ithought,Imusthavemadea
mistakewhowouldeverbuildacircuitthatbig???Well,Itoldtheinterviewerat
ElliottBrosaboutthedesignandlearnedthatyes,900wasactuallysmallbytheir
standards,andthearchitectureIhadchosenwasexactlyhowtheirprevious
generation803computersworks.Flooredme.AndIgotthejob.Neverlooked
back,stilllearningandinventingnearly50yearslater.Ievenbuiltthatcircuitandit
worked.
6 Reply Share

KonradHinsen 2daysago
Aniceperspectiveonmathsseenfromasoftwareengineeringpointofview,
thanks!

Inthediscussionofthenumbersystem,Iproposetoaddtwolayersinbetween
the"rational"andthe"real"numbers,whichtosomedegreeaddresstheproblems
thatarediscussed.

Thefirstlayerarethealgebraicnumbers,definedasthezerosofanypolynomial
withrationalcoefficients.Theytakecareofmanygeometricproblems,suchasthe
hypotenuseofarighttriangle,withoutgettingintotheinfinitytroublesofthe"full"
realnumbers.Thesetofalgebraicnumbersisstillcountable,whichcanbe
deducedfromthefactthatthenumberofpolynomialswithrationalcoefficients
remainscoutablyinfinite.

Thesecondlayerarethecomputablenumbers,definedasallnumbers
computablebyaTuringmachine,asexplainedbyTuringinhisfamous1936
paper.Evenanumberwithaninfinitedecimalrepresentationcanbecomputable,
inthesensethatonewriteaprogramthatproducesthecorrespondingstreamof
digitsasfarasdesired.ForanygivennumberNofdigits,thecomputation
seemore

2 Reply Share

AliceZheng>KonradHinsen adayago
Thankyouforthesuggestion.Iagreethatthesetwosubclassesarevery
relevant.I'dthoughtaboutmentioningcomputablenumbers,butwasn't
surehowmanypeoplewouldknowabouttheconcept.SoIdroppeditfor
thesakeofbeingsimpleandconcise.Butitiscertainlyavitallinkto
computerscience.Rather,itisthe"machineview"ofrealnumbers.

Theconceptofalgebraicnumbersisnewtome.Thanksforteachingme
somethingnewtoday!
1 Reply Share

Mathiness 3daysago
Interestingcorrelation.I'venotseentherealnumbersystemshowingrational
numbersasasubsetofirrationalnumbers.I'mnotsureifIunderstandhowthatbit
fits.
2 Reply Share

Jack>Mathiness 3daysago
Therationalsare*not*asubsetoftheirrationals,astheyare
complementary.Thediagramwouldbeconsistentif"irrational"was
replacedwith"real".
2 Reply Share

AliceZheng>Mathiness 3daysago
Yes,youarecorrect.Thanksfornoticingthebug.Thisshouldbefixed
soon.
1 Reply Share

MaxPhoton 3daysago
Ireallyappreciateyourpost.Andwhatafuncoincidence!Ijustposteda
fascinatinginterviewwithProf.AntalFekete,authorofauniquetext:RealLinear
Algebra.Ithinkyouandyourreaderswillfindtheinterviewintriguingandfullof
insight.Search"RealLinearAlgebra"orvisit:http://www.maxphoton.com/conve...
http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 5/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

insight.Search"RealLinearAlgebra"orvisit:http://www.maxphoton.com/conve...
2 Reply Share

PaulTopping 3daysago
Nicepost!Butdoesn'tmostofoursocietythinksoftwareengineeringisalso"too
abstract,tooimpenetrable,toodifficulttodigestandimpossibletoknow?"While
thedemandforsoftwareengineersisthehighestever,thenumberofCSdegrees
conferredinUScollegesisnotgoingupatnearthesamerate.
2 Reply Share

AnyaArisohn>PaulTopping adayago
Alothastodowithpeoplelearningfromothersintheprograming/software
industryoronlineresearchaboutcertificationinspecificcodingareas.A
CSdegreewillgetyoujustthat,adegreebutacertwillgetyoumuch
fartherandfaster,ex.UNIX,Cisco,C++,etc.themoreyouget,themore
you'llmake,andthefasteryou'llbehired:)
...IMO,gettheCSlateronifyou'rethinkingaboutadvancingtoupper
management,orwhateverbullsh!tdegreeworksfortheco.,like
management.lol
Reply Share

PaulTopping>AnyaArisohn adayago
Inmyworld,certificationsarebitsofpaperthatprovenothingmore
thanyouwereabletositthroughaclassyoupaidalottoattend.
CertificationsforITanddatabaseadmins,notrealprogrammers.
Reply Share

JoeMcCarthy 3daysago
Aninteresting&usefulperspective.Willwebeseeinganelaborationonthese
ideas,say,intheformofa"HeadFirstMathematics"or"HeadFirstMathematical
DesignPatterns"?
2 Reply Share

AnyaArisohn>JoeMcCarthy adayago
ThecurrentdirectiontheeducationsystemintheUSheaded,notanytime
soonmyfriend.:(
Reply Share

Noxyx adayago
Yawn.Anyonewhousesthephrase,"crosspollinateideas,"andwritesa1,200
wordessayjusttosay"mathandprogramminghavesomesimilarities,"probably
doesn'tbelonginanengineeringpositioninthefirstplace.
1 Reply Share

AnyaArisohn>Noxyx adayago
Asanelectricalengineermyself,IL0L'edatthattoo.Buthey,atleast
someoneonthe'outside'istryingtoexplainit...Hopefullythecross
pollinatepeoplewillspreadthegoodword!
Reply Share

MortenBrodersen adayago
Excellentarticle.Andyoumightbeevenmorerightthanyourealise.Rightnowa
revolutionistakingplaceinmathematics.TheideabeingthatTypeTheory,
CategoryTheoryandProofTheoryareinsomesenseequivalent.Inotherwords,
thatmathisfundamentallyaboutalgorithms(computations).Asanexample:the
axiomsofrationalandirrationalnumbersarealgorithms.Socomputerscienceis
notjustabranchofmath.ItISmathataverydeepfundamentallevel.
1 Reply Share

AliceZheng>MortenBrodersen adayago
That'sfascinating.Thanksfortheinfo!Idon'tknowanythingaboutType
Theory,CategoryTheory,orProofTheory.NeitherdoIhaveagoodgrasp
oftheformalmathematicalfoundationsofprogramminglanguages.So
whileI'mawareofthatintimateconnection,Idon'tknowenoughtouseitas
anexamplehere,unfortunately.

Astoyoursecondpointaboutrationalandirrationalnumbersbeing
algorithms,areyoureferringtoTuringcomputablenumbers,asKonrad
http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 6/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

algorithms,areyoureferringtoTuringcomputablenumbers,asKonrad
pointedoutbelow?Theconnectionbetweencomputerscience(notjust
softwareengineering)andmathematicsisindeedbidirectionaland
complexandfascinatingtonoend.

Thereareotherareasofmathematicsthataremorealgorithmlike.For
instance,thetwodefinitionsoftheintegralRiemannandLebesgueare
essentiallytwodifferentalgorithmsforcomputingthesamequantity.
Whereasfiniteprecisionisthebaneofcomputation,discontinuities(or
moregenerally,pathelogicaltopologies)arethebaneofanalysis.
Lebesgueintegralcircumventstheproblemofdiscontinuitiesthatcaused
problemsfortheRiemannintegral.

Anyway,thereissomuchmoretothesubject.Thanksforyourcomments!
1 Reply Share

MortenBrodersen>AliceZheng adayago
YouarewelcomeAlice:)Andyesitisadeepandfascinating
subjectindeed!Iamnotevenclosetobeinganexpertonthe
subject.Butlearningfunctionalprogramming(viaHaskell)opened
myeyestohowtrulyusefulmathisasatooltounderstandand
writesoftwarebetter.Theideabehind"mathasalgorithms"started
withBrouwerandConstructivism.Wikipediahasagreatarticle
aboutit.ThenextmajorstepwasMartinLoffTypeTheory.Wikiis
againagoodstartingpoint.Andontopofthatwenowhave
HomotopyTypeTheory.Itaimstodefineandunifyallofmathusing
machinecheckablealgorithms.Letmerepeatforemphasis:Allof
math.Usingalgorithms.Machinecheckable.Thatistrulymind
blowing:)Andbythewayanotherstartingpointisthesocalled
"CurryHowardcorrespondence".Itisthesameidea.Thatmath
andmathematicalproofsare"thesame"aswritingaprogram.
1 Reply Share

dorothyhs adayago
Thanks,thisisexcellent
1 Reply Share

simpleservant 3daysago
Well,softwarelanguagesweredesignedbymathematicians...It'sallcomplexity
conservation.Verynicearticle,hopefullyitwillgetmorepeopletoseethe
applicabilityofthinkingmathematically.
1 Reply Share

PaulTopping>simpleservant adayago
Mostprogramminglanguagesweredesignedbycomputerscientists,not
mathematicians.PerhapsintheearlydaysofcomputingwithFortranand
LISPbutnonesincethatcomestomind.
Reply Share

BenjaminCurrie 3daysago
Thanksforthepostthat'sapowerfulideawellexplained.Itappearsyoursecond
imageshowingthefieldringgroupinheritanceisinreverse?
1 Reply Share

AliceZheng>BenjaminCurrie 3daysago
Youareright."Group"and"Field"shouldbereversed.Ijustsentoffan
updatedillustrationanditshouldbecorrectedsoon.Thanksforcatching
thebug!
Reply Share

PaulTopping>AliceZheng adayago
Oneperson'sgroupisanotherperson'sfield.Orisittheotherway
around?
Reply Share

WilliamStein>BenjaminCurrie 3daysago
Iagreewithyou.It'sreversed,whichmaybeconfusingforreaders.
Reply Share
http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 7/8
22/1/2015 ParalelismossorprendentesentrelasmatemticasylaingenieradesoftwareO'ReillyRadar

Reply Share

MajidalDosari 12hoursago
unfortunately,thewaymathematicsisusuallytaughtinschoolis:
1.hereisthetheory.knowtheserulesandprocedures
2.doproblems15byfriday.
3.quiznextweek

thisconventionalwayofteachingmathdoesNOTattempttoshowthebeautyand
eleganceofmath.veryfewofmyinstructorsanswerthequestion:whyisthis
theorycool?

also,theabilitytousecomputerstoexperimentwithmathisunderutilized.

/rant
Reply Share

ALSOONO'REILLYRADAR WHAT'STHIS?

TheInternetofThingshasfourbig Selfcensorshipisnottheway
dataproblems forward
10comments10daysago 1comment10daysago
Huey"Ifdatadoesntchangeyour deemerySuggestyougobackand
behavior,whybothercollectingit?" rereadMarco's"retraction."Icertainly
Thedatayou'recollectingmight didnotperceivethatas

Fourshortlinks:8January2015 Fourshortlinks:26December2014
1comment15daysago 2commentsamonthago
jsonmez HeyNat,Thanksforthe ThomasBjelkemanIfbadsoftware
mentiontomypostondeveloper leadstobadscience,thenIammore
podcast.Appreciateit. worriedaboutbadstatistic.

Subscribe d AddDisqustoyoursite Privacy

http://radar.oreilly.com/2015/01/strikingparallelsbetweenmathematicsandsoftwareengineering.html 8/8

Potrebbero piacerti anche