FAQ

Imibuzo Ebuzwa Rhoqo malunga nedatha yokwenziwa

Iyaqondakala! Ngethamsanqa, sineempendulo kwaye silapha ukunceda. Jonga imibuzo yethu edla ngokubuzwa.

Nceda uvule umbuzo ongezantsi kwaye ucofe amakhonkco ukufumana ulwazi oluthe kratya. Ngaba unombuzo onzima ngakumbi ongachazwanga apha? Buza iingcali zethu ngokuthe ngqo!

Eyona mibuzo ibuzwayo

Idatha yeSynthetic ibhekisa kwidatha eyenziwe ngobuchule endaweni yokuba iqokelelwe kwimithombo yokwenyani yehlabathi. Ngokubanzi, ngelixa idatha yoqobo iqokelelwa kulo lonke unxibelelwano lwakho nabantu (abaxumi, izigulane, njl.) kwaye ngokusebenzisa zonke iinkqubo zakho zangaphakathi, idatha yokwenziwa iveliswa yi-algorithm yekhompyutha.

Idatha ye-synthetic ingasetyenziselwa ukuvavanya kunye nokuvavanya imodeli kwindawo elawulwayo, okanye ukukhusela ulwazi olubuthathaka ngokuvelisa idatha efana nedatha yehlabathi lokwenyani kodwa ingenalo naluphi na ulwazi olunovakalelo. Idatha yokwenziwa isoloko isetyenziswa njengenye yedatha ebuthathaka yabucala kwaye inokusetyenziswa njengedatha yovavanyo, kuhlalutyo okanye ukuqeqesha ukufunda koomatshini.

Funda ngokugqithisileyo

Ukuqinisekisa ukuba idatha yokwenziwa ibamba umgangatho wedatha efanayo njengedatha yoqobo kunokuba ngumngeni, kwaye kaninzi ixhomekeke kwimeko yokusetyenziswa kunye neendlela ezisetyenziselwa ukuvelisa idatha yokwenziwa. Ezinye iindlela zokuvelisa idatha yokwenziwa, njengemizekelo yokuvelisa, inokuvelisa idatha efana kakhulu nedatha yokuqala. Umbuzo ongundoqo: ungayibonisa njani le nto?

Kukho iindlela ezithile zokuqinisekisa umgangatho wedatha yokwenziwa:

  • Iimetriki zomgangatho wedatha ngengxelo yethu yobulunga bedatha: Enye indlela yokuqinisekisa ukuba idatha yokwenziwa ibamba umgangatho wedatha efanayo njengedatha yoqobo kukusebenzisa imilinganiselo yomgangatho wedatha ukuthelekisa idatha yokwenziwa kwidatha yokuqala. Ezi metrics zingasetyenziselwa ukulinganisa izinto ezifana nokufana, ukuchaneka, kunye nokugqibelela kwedatha. I-software ye-Syntho iquka ingxelo yomgangatho wedatha kunye neemetrices ezahlukeneyo zomgangatho wedatha.
  • Uvavanyo lwangaphandle: ekubeni umgangatho wedatha yedatha yokwenziwa xa kuthelekiswa nedatha yangaphambili ingundoqo, kutshanje senze uvavanyo kunye neengcali zedatha ye-SAS (inkokeli yemarike kwi-analytics) ukubonisa umgangatho wedatha yedatha yokwenziwa nguSyntho ngokuthelekisa idatha yangempela. U-Edwin van Unen, ingcali yohlalutyo evela kwa-SAS, uvavanye iiseti zedatha ezenziweyo ezivela kwi-Syntho ngokusebenzisa uhlalutyo olwahlukeneyo (AI) kunye nokwabelana ngeziphumo. Bukela isishwankathelo esifutshane saloo vidiyo apha.
  • Uvavanyo novavanyo uwedwa: idatha yokwenziwa ingavavanywa kwaye ivavanywe ngokuyithelekisa nedatha yehlabathi langempela okanye ngokuyisebenzisa ukuqeqesha iimodeli zokufunda ngomatshini kunye nokuthelekisa ukusebenza kwazo kwiimodeli eziqeqeshwe kwidatha yehlabathi langempela. Kutheni ungavavanyi umgangatho wedatha yedatha eyenziwe ngokwakho? Buza iingcali zethu malunga namathuba oku apha

Kubalulekile ukuqaphela ukuba idatha eyenziweyo ayinakuze iqinisekise ukuba i-100% ifana nedatha yokuqala, kodwa ingaba kufutshane ngokwaneleyo ukuba ibe luncedo kwimeko ethile yokusetyenziswa. Le meko ikhethekileyo yokusetyenziswa inokuba luhlalutyo oluphambili okanye imodeli yokufunda koomatshini boqeqesho.

Ukudala 'ukungaziwa' akusoloko kusesona sisombululo silungileyo, kuba:

  1. Umngcipheko wabucala – uya kuhlala unayo
    umngcipheko wabucala. Ukusebenzisa ezo
    ubuchule bokuzichaza
    yenza kube nzima kuphela, kodwa hayi
    akunakwenzeka ukuchonga abantu.
  2. Ukonakalisa idatha – ngakumbi wena
    ungazichazi amagama, kokukhona ukhusela
    ubumfihlo bakho, kodwa ngakumbi wena
    tshabalalisa idatha yakho. Ayisiyiyo le nto
    ufuna uhlalutyo, kuba
    data echithiweyo iya kubangela ububi
    ukuqonda.
  3. Itya ixesha – sisisombululo
    oko kuthatha ixesha elininzi, kuba
    ezo ndlela zisebenza ngokwahlukileyo
    ngokweseti yedatha nangohlobo lwedatha.

Idatha ye-Synthetic ijolise ekusombululeni zonke ezi ntsilelo. Umahluko uyamangalisa kangangokuba senze ividiyo ngayo. Jonga apha.

Imibuzo ebuzwa qho

Idatha yokwenziwa

Ngokubanzi, uninzi lwabathengi bethu basebenzisa idatha yokwenziwa:

  • Uvavanyo lweSoftware nophuhliso
  • Idatha yeSynthetic yohlalutyo, uphuhliso lwemodeli kunye nohlalutyo oluphambili (AI & ML)
  • Iidemo zemveliso

Funda ngakumbi kwaye ujonge iimeko zokusetyenziswa.

Iwele ledatha yokwenziwa yi-algorithm-eveliswa yikopi yedatha yelizwe lokwenyani kunye / okanye idatabase. Nge-Synthetic Data Twin, i-Syntho ijolise ekulinganiseni i-database yoqobo okanye isiseko sedatha ngokusondeleyo kangangoko kunokwenzeka kwidatha yentsusa ukwenza umboniso wokwenene woqobo. Ngewele ledatha yokwenziwa, sijonge umgangatho ophezulu wedatha eyenziweyo xa kuthelekiswa nedatha yokuqala. Oku sikwenza ngesoftware yethu yokwenziwa kwedatha esebenzisa imodeli ye-AI yesimanje. Ezo modeli ze-AI zivelisa iindawo zedatha ezintsha ngokupheleleyo kwaye ziyimodeli ngendlela yokuba sigcine iimpawu, ubudlelwane kunye neepateni zeenkcukacha-manani zedatha yoqobo ukuya kuthi ga kumlinganiselo wokuba ungayisebenzisa ngathi-ukuba yidatha yoqobo.

Oku kungasetyenziselwa iinjongo ezahlukeneyo, ezifana nokuvavanya kunye nokuqeqeshwa kweemodeli zokufunda zoomatshini, ukulinganisa iimeko zophando kunye nophuhliso, kunye nokudala iimeko ezibonakalayo zoqeqesho kunye nemfundo. I-synthetic data twins ingasetyenziselwa ukudala idatha yokwenyani kunye nommeli enokuthi isetyenziswe endaweni yedatha yelizwe lokwenyani xa ingafumaneki okanye xa usebenzisa idatha yelizwe lokwenyani ayinakwenzeka okanye ingabikho ngenxa yemithetho engqongqo yabucala.

Funda ngokugqithisileyo.

Ewe sakwenza njalo. Sinikezela ngexabiso elongezelelekileyo lokwenziwa kwedatha yokwenziwa kunye neempawu zokwandisa, kubandakanywa nabagculeli, ukuthatha idatha yakho ukuya kwinqanaba elilandelayo.

Funda ngokugqithisileyo.

Idatha ye-Mock kunye nedatha eyenziwe yi-AI-eveliswa yi-AI zombini iindidi zedatha yokwenziwa, kodwa ziveliswa ngeendlela ezahlukeneyo kwaye zikhonza iinjongo ezahlukeneyo.

Idatha ye-Mock luhlobo lwedatha yokwenziwa eyenziwe ngesandla kwaye isoloko isetyenziselwa uvavanyo kunye neenjongo zophuhliso. Ngokuqhelekileyo isetyenziselwa ukulinganisa ukuziphatha kwedatha yehlabathi lokwenyani kwindawo elawulwayo kwaye ihlala isetyenziselwa ukuvavanya ukusebenza kwenkqubo okanye isicelo. Ihlala ilula, kulula ukuyivelisa, kwaye ayifuni iimodeli ezintsonkothileyo okanye i-algorithms. Rhoqo, abanye ababhekisi baphinda bahlekise idatha njenge "data dummy" okanye "data yobuxoki".

I-AI-generated data synthetic data, ngakolunye uhlangothi, iveliswa kusetyenziswa ubuchule bobukrelekrele bokwenziwa, njengokufunda ngomatshini okanye iimodeli zokuvelisa. Isetyenziselwa ukwenza idatha yokwenyani kunye nemeleyo enokusetyenziswa endaweni yedatha yelizwe lokwenyani xa kusetyenziswa idatha yelizwe lokwenyani ayinakwenzeka okanye ingabikho ngokweenqobo ezisesikweni ngenxa yemigaqo engqongqo yabucala. Ihlala intsonkothe ​​ngakumbi kwaye ifuna izixhobo zokubala ngakumbi kunedatha ehlekisayo eyenziwe ngesandla. Ngenxa yoko, iyinyani ngakumbi kwaye ilinganisa idatha yoqobo ngokusondele kangangoko.

Isishwankathelo, idatha ehlekisayo yenziwa ngesandla kwaye iqhele ukusetyenziselwa uvavanyo kunye nophuhliso, ngelixa idatha eyenziwe yi-AI eyenziwe kusetyenziswa ubuchule bobukrelekrele bokwenziwa kwaye isetyenziselwa ukwenza idatha emele kunye neyenyani.

Imibuzo engakumbi? Buza iingcali zethu

Umgangatho weDatha

Ukuqinisekisa ukuba idatha yokwenziwa ibamba umgangatho wedatha efanayo njengedatha yoqobo kunokuba ngumngeni, kwaye kaninzi ixhomekeke kwimeko yokusetyenziswa kunye neendlela ezisetyenziselwa ukuvelisa idatha yokwenziwa. Ezinye iindlela zokuvelisa idatha yokwenziwa, njengemizekelo yokuvelisa, inokuvelisa idatha efana kakhulu nedatha yokuqala. Umbuzo ongundoqo: ungayibonisa njani le nto?

Kukho iindlela ezithile zokuqinisekisa umgangatho wedatha yokwenziwa:

  • Iimetriki zomgangatho wedatha ngengxelo yethu yobulunga bedatha: Enye indlela yokuqinisekisa ukuba idatha yokwenziwa ibamba umgangatho wedatha efanayo njengedatha yoqobo kukusebenzisa imilinganiselo yomgangatho wedatha ukuthelekisa idatha yokwenziwa kwidatha yokuqala. Ezi metrics zingasetyenziselwa ukulinganisa izinto ezifana nokufana, ukuchaneka, kunye nokugqibelela kwedatha. I-software ye-Syntho iquka ingxelo yomgangatho wedatha kunye neemetrices ezahlukeneyo zomgangatho wedatha.
  • Uvavanyo lwangaphandle: ekubeni umgangatho wedatha yedatha yokwenziwa xa kuthelekiswa nedatha yangaphambili ingundoqo, kutshanje senze uvavanyo kunye neengcali zedatha ye-SAS (inkokeli yemarike kwi-analytics) ukubonisa umgangatho wedatha yedatha yokwenziwa nguSyntho ngokuthelekisa idatha yangempela. U-Edwin van Unen, ingcali yohlalutyo evela kwa-SAS, uvavanye iiseti zedatha ezenziweyo ezivela kwi-Syntho ngokusebenzisa uhlalutyo olwahlukeneyo (AI) kunye nokwabelana ngeziphumo. Bukela isishwankathelo esifutshane saloo vidiyo apha.
  • Uvavanyo novavanyo uwedwa: idatha yokwenziwa ingavavanywa kwaye ivavanywe ngokuyithelekisa nedatha yehlabathi langempela okanye ngokuyisebenzisa ukuqeqesha iimodeli zokufunda ngomatshini kunye nokuthelekisa ukusebenza kwazo kwiimodeli eziqeqeshwe kwidatha yehlabathi langempela. Kutheni ungavavanyi umgangatho wedatha yedatha eyenziwe ngokwakho? Buza iingcali zethu malunga namathuba oku apha

Kubalulekile ukuqaphela ukuba idatha eyenziweyo ayinakuze iqinisekise ukuba i-100% ifana nedatha yokuqala, kodwa ingaba kufutshane ngokwaneleyo ukuba ibe luncedo kwimeko ethile yokusetyenziswa. Le meko ikhethekileyo yokusetyenziswa inokuba luhlalutyo oluphambili okanye imodeli yokufunda koomatshini boqeqesho.

Ewe yiyo. Idatha yokwenziwa ide ibambe iipateni obungazi ukuba bezikhona kwidatha yoqobo.

Kodwa musa nje ukuthatha ilizwi lethu ngayo. Iingcali ze-analytics ze-SAS (inkokeli yemarike yehlabathi kwi-analytics) yenze (AI) uhlolo lwedatha yethu yokwenziwa kwaye ithelekisa kunye nedatha yokuqala. Ngaba unomdla? Jonga i umsitho wonke apha okanye ubukele inguqulelo emfutshane malunga umgangatho wedatha apha.

Ewe sakwenza njalo. Iqonga lethu lenzelwe ugcino-lwazi kwaye ngenxa yoko, ukugcinwa kwemfezeko yereferensi phakathi kweeseti zedatha kwidatgabase.

Unomdla wokwazi ngakumbi ngale nto?

Buza iingcali zethu ngokuthe ngqo.

Ukuba bucala

Hayi asiyenzi. Sinokubeka ngokulula i-Syntho Injini kwindawo okanye kwilifu lakho labucala nge-docker.

Hayi. Siye salungisa iqonga lethu ngendlela enokuthi isetyenziswe ngokulula kwindawo ethembekileyo yomthengi. Oku kuqinisekisa ukuba idatha ayisoze ishiye indawo ethembekileyo yomthengi. Izinketho zokuthunyelwa kwendawo ethembekileyo yomthengi "kwindawo" kwaye "kwindawo yelifu yomthengi (ilifu labucala)".

Ukuzikhethela: I-Syntho ixhasa inguqulelo ebanjwe "kwilifu le-Syntho".

Hayi I-Syntho Engine yi-self-service platform. Ngenxa yoko, ukuvelisa idatha yokwenziwa kunye ne-Syntho Engine inokwenzeka ngendlela yokuba end-to-end inkqubo, u-Syntho akakwazi ukubona kwaye akaze afuneke ukuba aqhube idatha.

Ewe sikwenza oku ngengxelo yethu ye-QA.

 

Xa udityaniswa iseti yedatha, kubalulekile ukubonisa ukuba umntu akakwazi ukuphinda achonge abantu. Kwi le vidiyo, UMarijn wazisa imilinganiselo yabucala ekwingxelo yethu yobulunga ukubonisa oku.

Ingxelo ye-Syntho's QA iqulethe ezintathu umgangatho woshishino iimetrics zokuvavanya ubumfihlo bedatha. Uluvo ngasemva kwezi metrics nganye ngolu hlobo lulandelayo:

  • Idatha yokwenziwa (S) iya kuba “kufutshane kangangoko”, kodwa “ingkufutshane kakhulu” kwidatha ekujoliswe kuyo (T).
  • Idata ekhethwe ngokungakhethiyo (H) imisela ibhenchmark "ekufutshane kakhulu".
  • A isisombululo esifanelekileyo yenza idatha entsha eyenziweyo eziphatha kanye njengedatha yoqobo, kodwa engazange ibonwe ngaphambili (= H).

Enye yeemeko zokusetyenziswa ezigxininiswe ngokukodwa yi-Dutch Data Protection Authority isebenzisa idatha yokwenziwa njengedatha yovavanyo.

Okunye kunokufumaneka kweli nqaku.

Injini yeSyntho

I-Syntho Engine ithunyelwa kwi-Docker container kwaye inokuthunyelwa ngokulula kwaye ifakwe kwindawo oyithandayo.

Iinketho zokusasazwa ezinokwenzeka ziquka:

  • Kuqala
  • Naliphi na ilifu (labucala).
  • Nayiphi na enye indawo engqongileyo

Funda ngokugqithisileyo.

I-Syntho ikuvumela ukuba udibanise ngokulula kunye nedatha yakho, izicelo, imibhobho yedatha okanye iinkqubo zefayile. 

Sixhasa iziqhagamshelo ezahlukeneyo ezidityanisiweyo ukuze ukwazi ukunxulumana nomthombo-indalo (apho kugcinwa khona idatha yoqobo) kunye nendawo ekuyiwa kuyo (apho ufuna ukubhala khona idatha yakho eyenziweyo) ukwenzela end-to-end indlela edibeneyo.

Iimpawu zoqhagamshelo esizixhasayo:

  • Iplagi kwaye udlale ngeDocker
  • 20+ iziqhagamshelo zedathabheyisi
  • 20+ iziqhagamshelo zenkqubo yeefayile

Funda ngokugqithisileyo.

Ngokwemvelo, ixesha lesizukulwana lixhomekeke kubungakanani besiseko sedatha. Ngokwe-avareji, itheyibhile eneerekhodi ezingaphantsi kwe-1 yezigidi zenziwa phantsi kwemizuzu emi-5.

I-algorithms yokufunda koomatshini kaSyntho inokwenza ngcono izinto ngokubanzi kunye neerekhodi ezininzi ezikhoyo, ezinciphisa umngcipheko wabucala. Ubuncinci bomlinganiselo wekholomu kumqolo we-1:500 uyacetyiswa. Umzekelo, ukuba umthombo wetafile unezintlu ezi-6, kufuneka iqulathe ubuncinci bemiqolo engama-3000.

Hayi akunjalo. Nangona kungathatha umzamo othile ukuqonda ngokupheleleyo izibonelelo, ukusebenza kunye nokusetyenziswa kwamatyala edatha eyenziweyo, inkqubo yokudibanisa ilula kakhulu kwaye nabani na onolwazi olusisiseko lwekhompyuter angayenza. Ngolwazi oluthe kratya malunga nenkqubo yokudibanisa, khangela eli phepha or ucele idemo.

I-Syntho Engine isebenza kakuhle kwi-structured, data tabular (nantoni na equlethe imigqa kunye neekholamu). Ngaphakathi kwezi zakhiwo, sixhasa ezi ntlobo zedatha zilandelayo:

  • Idatha yolwakhiwo efomathiweyo kwiitheyibhile (ngokwecandelo, ngokwamanani, njl.njl.)
  • Izazisi ezithe ngqo kunye nePII
  • Iiseti zedatha ezinkulu kunye nogcino-lwazi
  • Idatha yendawo yejografi (njengeGPS)
  • Idatha yochungechunge lwexesha
  • Ugcino lwedatha yeetheyibhile ezininzi (kunye nemfezeko yereferensi)
  • Vula idatha yombhalo

 

Inkxaso yedatha enzima
Ecaleni kwazo zonke iintlobo eziqhelekileyo zedatha ye-tabular, i-Syntho Engine isekela iintlobo zedatha eziyinkimbinkimbi kunye nezakhiwo zedatha eziyinkimbinkimbi.

  • Uthotho lwamaxesha
  • Uluhlu lweenkcukacha ezininzi
  • Vula umbhalo

Funda ngokugqithisileyo.

Hayi, siphucule iqonga lethu ukuze sinciphise iimfuno zokubala (umzekelo, akukho GPU ifunekayo), ngaphandle kokubeka esichengeni ukuchaneka kwedatha. Ukongeza, sixhasa ukulinganisa okuzenzekelayo, ukwenzela ukuba umntu akwazi ukuhlanganisa i-database enkulu.

Ewe. Isoftware yeSyntho ilungiselelwe idatabase equlathe iitafile ezininzi.

Ngokubhekiselele koku, i-Syntho ibona ngokuzenzekelayo iindidi zedatha, i-schemas kunye neefomathi zokwandisa ukuchaneka kwedatha. Kwisiseko sedatha yeetheyibhile ezininzi, sixhasa intelekelelo yobudlelwane betafile ngokuzenzekelayo kunye nokudibanisa ukugcina intembeko yereferensi.

iqela labantu abancumileyo

Idatha yenziwe, kodwa iqela lethu liyinyani!

Nxibelelana noSyntho kwaye enye yeengcali zethu iya kunxibelelana nawe ngesantya sokukhanya ukuphonononga ixabiso ledatha eyenziweyo!