Ixesha lokufunda eliqikelelweyo: 6 imizuzu
Iinkampani zivelisa amanani ahlala enyuka edatha ehambelana nemisebenzi yoshishino, ekhokelela kumdla ohlaziyiweyo kwi-analytics yokuxela kwangaphambili, intsimi ehlalutya iiseti zedatha enkulu ukuchonga iipatheni, ukuqikelela iziphumo kunye nokukhokela izigqibo. Iinkampani zikwajongene noluhlu oluntsonkothileyo nolusoloko lusanda lwemingcipheko yokusebenza ekufuneka ichongwe kwaye incitshiswe. Ngelixa iinkampani ezininzi ziye zaqala ukusebenzisa uhlalutyo lokuxela kwangaphambili ukuchonga amathuba okuthengisa / okuthengisa, izicwangciso ezifanayo aziqhelekanga kakhulu ekulawuleni umngcipheko, kubandakanya ukhuseleko.
Ii-algorithms zokuhlela, iklasi eliqhelekileyo lokuhlalutya kwangaphambili, linokuba luncedo ngokukodwa ekucoceni kunye namashishini e-petrochemical ngokuxela kwangaphambili ixesha kunye nendawo yeziganeko zokhuseleko ezisekelwe kwidatha yokuhlola kunye nokugcinwa kwedatha enxulumene nokhuseleko, ngokubalulekileyo izikhombisi ezihamba phambili. Kukho imingeni emibini engundoqo ehambelana nale ndlela: (1) ukuqinisekisa ukuba izikhombisi ezikhokelayo ezilinganisiweyo ngokwenene zixela kwangaphambili ukuphazamiseka kunye (2) nokulinganisa izikhombisi ezihamba phambili rhoqo ngokwaneleyo ukuba zibe nexabiso eliqikelelweyo.
Ukusebenzisa idatha yokuhlola ehlaziywa rhoqo, imodeli inokudalwa ngokusebenzisa ukuguqulwa kwezinto. Ngale ndlela ungadala imodeli, umzekelo, ukuqikelela ukusilela kololiwe kwimayile nganye yendlela. Okunokwenzeka kunokuhlaziywa njengoko idatha eyongezelelweyo iqokelelwa.
Ukongeza kumathuba aqikelelweyo okusilela kololiwe, ngemodeli efanayo sinokuchonga izinto eziguquguqukayo ezinobungqina obukhulu bokuqikelelwa (ezo zinegalelo elikhulu ekusileleni kukaloliwe). Ukusebenzisa iziphumo zemodeli, uya kukwazi ukuchonga ngokuthe ngqo apho ugxininise khona ukugcinwa, ukuhlolwa kunye nezixhobo zokuphucula inkunzi kunye nokuba yeyiphi imiba yokujongana nale misebenzi.
Kwale ndlela inye inokusetyenziswa kumashishini okucoca kunye nepetrochemical ukulawula imingcipheko ngokuqikelela nokuthintela iingozi, ngaphandle kokuba imibutho:
Uhlalutyo oluqikelelwayo yintsimi ebanzi ebandakanya iinkalo zezifundo ezahlukeneyo, kubandakanya ukufunda koomatshini,ingqiqo, izibalo kunye Imayini yedatha. Uhlalutyo oluqikelelweyo lutyhila iipateni kunye nentsingiselo kwiiseti zedatha enkulu. Olunye uhlobo lohlalutyo oluqikelelweyo, i-algorithms yokuhlelwa, lunokuba luncedo ngokukodwa kumashishini okucoca kunye ne-petrochemical.
Ii-algorithms zokuhlela zingahlelwa njengokufunda koomatshini okulawulwayo. Ngokufunda okugadwayo, umsebenzisi unedathasethi ebandakanya imilinganiselo yeenguqu eziqikelelwayo ezinokudityaniswa neziphumo ezaziwayo. Kumzekelo oxutyushwayo kwicandelo lokufunda eli nqaku, imilinganiselo eyahlukeneyo yomkhondo (umzekelo, i-curvature, i-crossing) ithathwe ngexesha le-mile nganye yomzila. Isiphumo esaziwayo, kulo mzekelo, kukuba ukusilela komzila kwenzeke kwimayile nganye kaloliwe ngelo xesha leminyaka mibini.
I-algorithm efanelekileyo yokulinganisa ikhethwa kwaye isetyenziselwa ukuhlalutya idatha kunye nokuchonga ubudlelwane phakathi kwemilinganiselo eguquguqukayo kunye neziphumo zokudala imithetho eqikelelwayo (umzekelo). Emva kokuba yenziwe, imodeli inikwa idatha entsha equlethe imilinganiselo yezinto eziguquguqukayo ezingaziwayo kunye neziphumo kwaye iya kubala ukubana kwesiphumo ngokusekelwe kwimigaqo yomzekelo. Oku kuthelekiswa neentlobo zokufunda okungagadwanga, apho i-algorithms ibona iipateni kunye nentsingiselo kwi-dataset ngaphandle kolwalathiso oluthile oluvela kumsebenzisi, ngaphandle kwe-algorithm esetyenzisiweyo.
Ii-algorithms zokuhlela eziqhelekileyo ziquka ukuhlehla komgca, ukuhlehliswa kwezinto, umthi wesigqibo, inethiwekhi ye-neural, i-vector yenkxaso / umatshini wokucalula oguquguqukayo, udidi lweBayes ongenalwazi, kunye nabanye abaninzi. Ukuhlehla komgca kunika umzekelo olula wendlela i-algorithm yokuhlelwa isebenza ngayo. Kulayini wokuhlehla, owona mgca ufanelekileyo ubalwa ngokusekelwe kumanqaku edatha asele ekhona, ukunika ukulingana komgca ay = mx + b. Ukufaka inguqu eyaziwayo (x) ibonelela ngengqikelelo yenguqu engaziwayo (y).
Ubudlelwane obuninzi phakathi kwezinto eziguquguqukayo kwihlabathi lokwenyani abukho mgca, kodwa buntsonkothile kwaye bumile ngendlela engaqhelekanga. Ngoko ke, ukuhlehla komgca kaninzi akuloncedo. Ezinye ii-algorithms zokuhlela ziyakwazi ukwenza imodeli yobudlelwane obunzima ngakumbi, obufana nobudlelwane be-curvilinear okanye i-logarithmic. Ngokomzekelo, i-algorithm yokubuyisela izinto inokwenza imodeli yobudlelwane obuntsonkothileyo, inokubandakanya okuguquguqukayo okungengamanani (umzekelo, iindidi), kwaye kaninzi inokudala imifuziselo eyinyani kunye neyokubalwa esebenzayo. Isiphumo esiqhelekileyo semodeli yokubuyisela umva yinto eqikelelweyo yesiphumo/isiganeko esenzekayo. Ezinye ii-algorithms zokuhlela zibonelela ngemveliso efanayo kwi-logistic regression, kodwa amagalelo afunekayo ahlukile phakathi kwe-algorithms.
Ukumodareyitha ubudlelwane obuntsonkothileyo buluncedo ngakumbi kulawulo lomngcipheko, apho umngcipheko ubekwe phambili ngokuqhelekileyo ngokusekelwe kumathuba kunye nobunzima obunokubakho besiphumo esithile. Ukwenza umzekelo wemingcipheko enegalelo kweso siphumo kubangela uqikelelo oluchanekileyo kunye noluchanekileyo lwezibalo lwamathuba esiphumo. Ngokwahlukileyo koko, iimvavanyo ezininzi zomngcipheko zilinganisa "ithuba" kwinqanaba le-categorical (kanye kwiminyaka elishumi, kanye ngonyaka, amaxesha amaninzi ngonyaka), engachanekanga kakhulu, i-subjective, kwaye yenza kube nzima ukwahlula phakathi kweengozi ezikhoyo kumngcipheko. udidi olubanzi olufanayo. Kukho ezinye iindlela zokuvavanya ubungqongqo obunokubakho kuvavanyo lomngcipheko, kodwa oku kungaphaya kwemida yeli nqaku.
BlogInnovazione.it
Icandelo lomkhosi wasemanzini ligunya lokwenyani loqoqosho lwehlabathi, elithe lajonga kwimarike ye-150 yeebhiliyoni...
NgoMvulo ophelileyo, i-Financial Times ibhengeze isivumelwano kunye ne-OpenAI. I-FT ikhupha iilayisensi kubuntatheli bayo obukumgangatho wehlabathi…
Izigidi zabantu zihlawula iinkonzo zokusasaza, zihlawula umrhumo wenyanga. Luluvo oluqhelekileyo ukuba…
I-Coveware nguVeeam iya kuqhubeka nokubonelela ngeenkonzo zokuphendula ngeziganeko zokuxhwilwa kwe-cyber. I-Coveware iya kubonelela nge-forensics kunye nezakhono zokulungisa…
