5.3.1 Netflix nqi zog

Lub Netflix nqi zog siv qhib kev hu mus rau kwv yees uas movies neeg yuav nyiam.

Qhov zoo tshaj plaws paub hu qhib qhov project yog Netflix nqi zog. Netflix yog ib lub yeeb yaj kiab hauv internet, thiab hauv 2000 nws tau qhib Cinematch, ib qhov kev pab cuam kom pom zoo rau cov neeg tuaj yeem. Piv txwv li, Cinematch yuav pom tias koj nyiam Star Wars thiab Lub Tsiaj Txhu Riam Tsov Rov Qab thiab tom qab ntawd koj xav kom koj saib Rov qab ntawm Jedi . Pib, Cinematch ua haujlwm tsis zoo. Tab sis, dhau ntawm lub xyoo ntau, nws tseem ua kom nws lub peev xwm los twv seb dab tsi cov neeg muas zaub yuav txaus siab. Los ntawm 2006, txawm li cas los, kev kawm tau ntawm Cinematch tau plag tsev. Cov neeg tshawb xyuas ntawm Netflix tau sim zoo nkauj npaum txhua yam lawv yuav xav txog, tab sis, tib lub sij hawm, lawv xav tias muaj lwm cov tswv yim uas yuav pab tau lawv ua kom lawv zoo zog. Yog li, lawv tuaj nrog dab tsi yog, thaum lub sij hawm, ib qho kev daws radical: qhib kev hu.

Tseem ceeb heev rau qhov kev tshwm sim tau zoo ntawm Netflix Prize yog li cas qhib hu raug tsim, thiab qhov tsim muaj qhov tseem ceeb ntawm kev hu xov tooj qhib yuav siv tau rau kev tshawb nrhiav. Netflix tsis cia li muab tso tawm ib qho kev thov kom tsis pub muaj tswv yim, uas yog cov neeg xav txog lawv thaum lawv thawj zaug qhib kev sib tham. Tiam sis, Netflix ua rau muaj qhov teeb meem tseeb uas yooj yim nrog txoj kev ntsuam xyuas yooj yooj yim: lawv tau twv kom cov neeg siv ib txheej ntawm 100 lab lub yeeb yam kev ntsuas kom pom tau 3 lab tawm-tawm kev (kev ntseeg tias cov neeg tau ua, tiam sis cov Netflix tsis tso tawm). Thawj tus neeg los tsim ib qho kev ntsuas uas tau pom tias 3 lab muaj kev tawm tsam 10% zoo tshaj Cinematch yuav yeej ib lab nyiaj. Qhov no meej thiab yooj yim rau kev siv cov txheej txheem ntsuam xyuas-piv tau cov kev ntsuam xyuas cov kev ntsuam xyuas uas tau muaj kev tawm-txhais tau hais tias Netflix Prize tau muab kho kom zoo li qhov kev daws tau yooj yim los xyuas dua li tsim tawm; nws muab cov kev sib tw ntawm kev txhim kho Cinematch rau hauv ib qho teeb meem uas tsim nyog rau kev qhib lus.

Nyob rau lub Kaum Hlis Ntuj xyoo 2006, Netflix tau tso tawm cov dataset uas muaj 100 lab lub yeeb yam los ntawm kwv yees li 500,000 tus neeg yuav khoom (peb yuav xav txog qhov tsis zoo ntawm cov ntaub ntawv tso tawm hauv tshooj 6). Netflix cov ntaub ntawv yuav conceptualized ua ib qho loj loj uas yog kwv yees li ntawm 500,000 tus qhauj los ntawm 20,000 tsos. Tom qab qhov ntsuas no, muaj txog 100 lab kev ntsuas ntawm ib teev mus rau tsib hnub qub (cov lus 5.2). Cov kev sib tw yog los siv cov ntaub ntawv raug cai hauv daim duab txhawm rau twv seb cov kev tawm tsam 3 lab puas tawm.

Cov lus 5.2: Schematic ntawm Cov Ntaub Ntawv ntawm Netflix
Yeeb yam 1 Yeeb yam 2 Yeeb yam 3 ... Yeeb yam 20,000
Neeg 1 2 5 ... ?
Neeg 2 2 ? ... 3
Neeg 3 ? 2 ...
\(\vdots\) \(\vdots\) \(\vdots\) \(\vdots\) \(\vdots\)
Neeg 500,000 ? 2 ... 1

Cov neeg tshawb nrhiav thiab cov neeg tshawb xyuas hackers thoob ntiaj teb raug coj los sib tw, thiab los ntawm 2008 ntau tshaj 30,000 tus neeg tau ua haujlwm rau nws (Thompson 2008) . Nyob rau hauv chav kawm ntawm lub contest, Netflix tau txais ntau tshaj 40,000 thov kev daws los ntawm ntau tshaj 5,000 pawg (Netflix 2009) . Obviously, Netflix tsis tuaj yeem nyeem thiab to taub tag nrho cov lus thov no. Tag nrho cov khoom tau khiav ncaj zoo, txawm li cas los, vim hais tias daws cov teeb meem yooj yim. Netflix yuav cia li muaj ib lub computer piv rau cov kev sib tw nrog cov txheej txheem uas tau muab siv ua ntej (siv cov metric lawv siv yog cov hauv paus hauv paus ntawm qhov phem sib txuam). Nws yog qhov kev muaj peev xwm ntsuam xyuas sai sai uas Netflix tau txais kev pom zoo los ntawm txhua tus neeg, uas tau hloov ua qhov tseem ceeb vim hais tias cov tswv yim zoo los ntawm tej qhov chaw uas xav tsis thoob. Qhov tseeb, txoj kev sib tw yeej tau xa los ntawm ib pab neeg tau pib los ntawm peb tug neeg soj ntsuam uas tsis muaj cov kev ua yeeb yaj kiab hauv tsev (Bell, Koren, and Volinsky 2010) .

Ib tug zoo nkauj nam ntawm Netflix nqi zog yog tias nws tau ua tag nrho cov lus thov los mus ntsuam xyuas ncaj. Qhov no yog, thaum cov neeg tau muab lawv qhov kev twv ua ntej, lawv tsis tas tso lawv cov ntaub ntawv kev kawm, lawv lub hnub nyoog, haiv neeg, poj niam txiv neej, kev nyiam sib deev, lossis txhua yam txog lawv tus kheej. Qhov kev twv ntawm ib tug xibfwb nto moo ntawm Stanford tau kho tau zoo tib yam li cov tub ntxhais hluas hauv nws chav pw. Hmoov tsis, qhov no tsis yog muaj tseeb nyob rau hauv kev tshawb fawb tshaj plaws. Ntawd yog, rau feem ntau cov kev tshawb fawb, kev ntsuam xyuas yog siv sij hawm heev thiab ib qho kev xav. Yog li, feem ntau cov tswv yim tshawb fawb yeej tsis raug ntsuas, thiab thaum twg cov tswv yim raug soj ntsuam, nws yuav nyuaj rau tshem tawm cov kev ntsuam xyuas los ntawm tus creator ntawm lub tswv yim. Qhib kev hu xovtooj qhib hu xov tooj rau lwm tus, muaj kev ntsuam xyuas tau yooj yim thiab ncaj ncees kom lawv paub txog cov tswv yim uas yuav tsis muaj lwm yam.

Piv txwv, ntawm ib tus taw tes thaum Netflix Prize, ib tug neeg uas muaj lub npe hu ua Simon Funk muab tso rau hauv nws cov lus tawm tswv yim los ntawm nws qhov kev sib txuas lus ntawm tus nqi, qhov kev tawm ntawm linear algebra uas tsis tau muab siv yav tas los los ntawm lwm cov neeg koom. Funk tus ncej blog yog ib txhij technical thiab weirdly kev. Yog no blog post piav txog ib qho kev daws teeb meem zoo lossis nws yog ib qho pov tseg ntawm lub sijhawm? Kev tawm sab nrauv ntawm qhov kev qhib hu, qhov kev daws teeb meem yuav tsis tau txais kev ntsuam xyuas loj. Tom qab tag nrho, Simon Funk tsis yog xibfwb hauv MIT; nws yog tus software tsim tawm, leej twg, thaum lub sij hawm, tau rov qab los ncig New Zealand (Piatetsky 2007) . Yog hais tias nws tau xa email no lub tswv yim mus rau tus engineer ntawm Netflix, nws yeej yuav tsis tau nyeem.

Qhov zoo ces, vim tias qhov kev ntsuam xyuas tau muaj tseeb thiab yooj yim rau kev thov, nws qhov kev ntsuam xyuas tau raug soj ntsuam, thiab nws tau paub meej meej tias nws txoj kev ua tau zoo heev: nws rocketed rau plaub qhov kev sib tw, ib qho kev tshwm sim zoo dua uas lwm pawg tau ua hauj lwm rau lub hlis ntawm qhov teeb meem. Thaum kawg, qhov chaw ntawm nws txoj kev ua tau raug siv los ntawm txhua tus sib tw loj heev (Bell, Koren, and Volinsky 2010) .

Qhov tseeb tias Simon Funk tau xaiv sau ntawv blog uas piav txog nws txoj kev qhia, es tsis txhob ua raws li nws daim ntawv qhia, nws kuj qhia tias ntau tus neeg koom hauv Netflix Qhov khoom plig tsis yog tshwj xeeb ntawm kev txhawb zog los ntawm tus nqi nyiaj them poob haujlwm lab. Tiam sis, ntau tus neeg tuaj koom kuj tau txais kev zoo siab thiab lub zej zog uas tsim los ntawm qhov teeb meem (Thompson 2008) , cov kev xav uas kuv xav kom ntau tus neeg tshawb xyuas yuav nkag siab.

Netflix Prize yog ib qho piv txwv ntawm kev qhib lus. Netflix posed ib lo lus nug nrog lub hom phiaj (twv saib cov yeeb yam) thiab cov ntsiab lus ntawm ntau tus neeg. Netflix tau soj ntsuam tag nrho cov kev daws teeb meem vim tias lawv tau yooj yim los xyuas dua li tsim, thiab thaum kawg Netflix khaws cov tshuaj zoo tshaj. Tom ntej no, kuv mam li qhia koj tias qhov kev qhia ntawv no yuav siv tau licas hauv biology thiab txoj cai lij choj, thiab tsis muaj ib qho nyiaj poob rau ib plhom-vam nyiaj.