Sei sulla pagina 1di 2

28/09/2015

ML:IntroductionCoursera

CourseraBeta
ANOOP

ContactUs
MyContributions
LogOut

ML:Introduction
FromCoursera

WhatisMachineLearning?
TwodefinitionsofMachineLearningareoffered.ArthurSamueldescribeditas:"thefieldofstudythatgives
computerstheabilitytolearnwithoutbeingexplicitlyprogrammed."Thisisanolder,informaldefinition.
TomMitchellprovidesamoremoderndefinition:"AcomputerprogramissaidtolearnfromexperienceEwith
respecttosomeclassoftasksTandperformancemeasureP,ifitsperformanceattasksinT,asmeasuredby
P,improveswithexperienceE."
Example:playingcheckers.
E=theexperienceofplayingmanygamesofcheckers
T=thetaskofplayingcheckers.
P=theprobabilitythattheprogramwillwinthenextgame.

SupervisedLearning
Insupervisedlearning,wearegivenadatasetandalreadyknowwhatourcorrectoutputshouldlooklike,
havingtheideathatthereisarelationshipbetweentheinputandtheoutput.
Supervisedlearningproblemsarecategorizedinto"regression"and"classification"problems.Ina
regressionproblem,wearetryingtopredictresultswithinacontinuousoutput,meaningthatwearetryingto
mapinputvariablestosomecontinuousfunction.Inaclassificationproblem,weareinsteadtryingtopredict
resultsinadiscreteoutput.Inotherwords,wearetryingtomapinputvariablesintodiscretecategories.
Example:
Givendataaboutthesizeofhousesontherealestatemarket,trytopredicttheirprice.Priceasafunctionof
sizeisacontinuousoutput,sothisisaregressionproblem.
Wecouldturnthisexampleintoaclassificationproblembyinsteadmakingouroutputaboutwhetherthehouse
"sellsformoreorlessthantheaskingprice."Hereweareclassifyingthehousesbasedonpriceintotwo
discretecategories.

UnsupervisedLearning
https://share.coursera.org/wiki/index.php/ML:Introduction

1/2

28/09/2015

ML:IntroductionCoursera

Unsupervisedlearning,ontheotherhand,allowsustoapproachproblemswithlittleornoideawhatourresults
shouldlooklike.Wecanderivestructurefromdatawherewedon'tnecessarilyknowtheeffectofthevariables.
Wecanderivethisstructurebyclusteringthedatabasedonrelationshipsamongthevariablesinthedata.
Withunsupervisedlearningthereisnofeedbackbasedonthepredictionresults,i.e.,thereisnoteacherto
correctyou.Itsnotjustaboutclustering.Forexample,associativememoryisunsupervisedlearning.
Example:
Clustering:Takeacollectionof1000essayswrittenontheUSEconomy,andfindawaytoautomatically
grouptheseessaysintoasmallnumberthataresomehowsimilarorrelatedbydifferentvariables,suchasword
frequency,sentencelength,pagecount,andsoon.
Associative:Supposeadoctoroveryearsofexperienceformsassociationsinhismindbetweenpatient
characteristicsandillnessesthattheyhave.Ifanewpatientshowsupthenbasedonthispatientscharacteristics
suchassymptoms,familymedicalhistory,physicalattributes,mentaloutlook,etcthedoctorassociatespossible
illnessorillnessesbasedonwhatthedoctorhasseenbeforewithsimilarpatients.Thisisnotthesameasrule
basedreasoningasinexpertsystems.Inthiscasewewouldliketoestimateamappingfunctionfrompatient
characteristicsintoillnesses.
Next:LinearRegressionwithOneVariable
BacktoIndex:Main
Retrievedfrom"https://share.coursera.org/wiki/index.php?title=ML:Introduction&oldid=18622"
Category:
ML:LectureNotes
Thispagewaslastmodifiedon7November2013,at15:23.
Thispagehasbeenaccessed41,075times.
Privacypolicy
AboutCoursera
Disclaimers

https://share.coursera.org/wiki/index.php/ML:Introduction

2/2

Potrebbero piacerti anche