Acknowledgments
IwouldliketothankmyGooglecolleagues,inparticulartheYouTubevideoclassificationteam,for
teachingmesomuchaboutMachineLearning.Icouldneverhavestartedthisprojectwithoutthem.
SpecialthankstomypersonalMLgurus:ClémentCourbet,JulienDubois,MathiasKende,Daniel
Kitachewsky,JamesPack,AlexanderPak,AnoshRaj,VitorSessak,WiktorTomczak,IngridvonGlehn,
RichWashington,andeveryoneatYouTubeParis.
Iamincrediblygratefultoalltheamazingpeoplewhotooktimeoutoftheirbusylivestoreviewmybook
insomuchdetail.ThankstoPeteWardenforansweringallmyTensorFlowquestions,reviewingPartII,
providingmanyinterestinginsights,andofcourseforbeingpartofthecoreTensorFlowteam.Youshould
definitelycheckouthisblog!ManythankstoLukasBiewaldforhisverythoroughreviewofPartII:he
leftnostoneunturned,testedallthecode(andcaughtafewerrors),mademanygreatsuggestions,andhis
enthusiasmwascontagious.Youshouldcheckouthisblogandhiscoolrobots!ThankstoJustinFrancis,
whoalsoreviewedPartIIverythoroughly,catchingerrorsandprovidinggreatinsights,inparticularin
Chapter16.CheckouthispostsonTensorFlow!
HugethanksaswelltoDavidAndrzejewski,whoreviewedPartIandprovidedincrediblyuseful
feedback,identifyingunclearsectionsandsuggestinghowtoimprovethem.Checkouthiswebsite!
ThankstoGrégoireMesnil,whoreviewedPartIIandcontributedveryinterestingpracticaladviceon
trainingneuralnetworks.ThanksaswelltoEddyHung,SalimSémaoune,KarimMatrah,Ingridvon
Glehn,IainSmears,andVincentGuilbeauforreviewingPartIandmakingmanyusefulsuggestions.AndI
alsowishtothankmyfather-in-law,MichelTessier,formermathematicsteacherandnowagreat
translatorofAntonChekhov,forhelpingmeironoutsomeofthemathematicsandnotationsinthisbook
andreviewingthelinearalgebraJupyternotebook.
Andofcourse,agigantic“thankyou”tomydearbrotherSylvain,whoreviewedeverysinglechapter,
testedeverylineofcode,providedfeedbackonvirtuallyeverysection,andencouragedmefromthefirst
linetothelast.Loveyou,bro!
ManythanksaswelltoO’Reilly’sfantasticstaff,inparticularNicoleTache,whogavemeinsightful
feedback,alwayscheerful,encouraging,andhelpful.ThanksaswelltoMarieBeaugureau,BenLorica,
MikeLoukides,andLaurelRumaforbelievinginthisprojectandhelpingmedefineitsscope.Thanksto
MattHackerandalloftheAtlasteamforansweringallmytechnicalquestionsregardingformatting,
asciidoc,andLaTeX,andthankstoRachelMonaghan,NickAdams,andalloftheproductionteamfor
theirfinalreviewandtheirhundredsofcorrections.
Lastbutnotleast,Iaminfinitelygratefultomybelovedwife,Emmanuelle,andtoourthreewonderful
kids,Alexandre,Rémi,andGabrielle,forencouragingmetoworkhardonthisbook,askingmany
questions(whosaidyoucan’tteachneuralnetworkstoaseven-year-old?),andevenbringingmecookies
andcoffee.Whatmorecanonedreamof?
AvailableonHinton’shomepageathttp://www.cs.toronto.edu/~hinton/.
DespitethefactthatYannLecun’sdeepconvolutionalneuralnetworkshadworkedwellforimagerecognitionsincethe1990s,although
theywerenotasgeneralpurpose.
1
2