[machine_learning_mastery系列]BasicsforLinearAlgebraforMachineLearning.pdf

[machine_learning_mastery系列]BasicsforLinearAlgebraforMachineLearning.pdf
PrefaceIwrotethisbooktohelpmachinelearningpractitioners,likeyou,getontopoflinearalgebra,fast.LinearAlgebraIsImportantinMachineLearningThereisnodoubtthatlinearalgebraisimportantinmachinelearning.Linearalgebraisthemathematicsofdata.It’sallvectorsandmatricesofnumbers.Modernstatisticsisdescribedusingthenotationoflinearalgebraandmodernstatisticalmethodsharnessthetoolsoflinearalgebra.Modernmachinelearningmethodsaredescribedthesameway,usingthenotationsandtoolsdrawndirectlyfromlinearalgebra.Evensomeclassicalmethodsusedinthefield,suchaslinearregressionvialinearleastsquaresandsingular-valuedecomposition,arelinearalgebramethods,andothermethods,suchasprincipalcomponentanalysis,werebornfromthemarriageoflinearalgebraandstatistics.Toreadandunderstandmachinelearning,youmustbeabletoreadandunderstandlinearalgebra.PractitionersStudyLinearAlgebraTooEarlyIfyouaskhowtogetstartedinmachinelearning,youwillverylikelybetoldtostartwithlinearalgebra.Weknowthatknowledgeoflinearalgebraiscriticallyimportant,butitdoesnothavetobetheplacetostart.Learninglinearalgebrafirst,thencalculus,probability,statistics,andeventuallymachinelearningtheoryisalongandslowbottom-uppath.Abetterfitfordevelopersistostartwithsystematicproceduresthatgetresults,andworkbacktothedeeperunderstandingoftheory,usingworkingresultsasacontext.Icallthisthetop-downorresults-firstapproachtomachinelearning,andlinearalgebraisnotthefirststep,butperhapsthesecondorthird.PractitionersStudyTooMuchLinearAlgebraWhenpractitionersdocirclebacktostudylinearalgebra,theylearnfarmoreofthefieldthanisrequiredfororrelevanttomachinelearning.Linearalgebraisalargefieldofstudythathastendrilsintoengineering,physicsandquantumphysics.Therearealso