Akungabazeki ukuthi umhlaba wezocwaningo uvame ukuletha izimanga njalo, ikakhulukazi ezihlobene nomkhakha okhombise ukuthi uyaphikisana impela maqondana nokufundiswa, njengokuthuthuka izinhlelo zobunhloli bokufakelwa. Ngalesi sikhathi kufanele sikhulume nomsebenzi weqembu labaphenyi abakwazile ukwenza uhlelo olukwaziyo ukuthi ucabangani nganoma yisiphi isikhathi.
Umqondo laba bososayensi abanabo wokuthi lolu hlelo lube ukwazi ukwazi ukuthi ucabangani Akukona okunye ukwenza ukuskena kobuchopho kumuntu noma esifundweni sokuhlola lapho kwenziwa khona ukucubungula okuphelele kobuchopho ukuze kuchazwe ubuchopho balo muntu. Akukho ukushoda kososayensi abathi, ngenxa yale ntuthuko ehehayo, isayensi izokwazi ukuqhubeka ithuthuke ngokwemigomo ye- ulwazi ngokusebenza kobuchopho uqobo.
Abaphenyi abavela ePurdue University bayakwazi ukwenza uhlelo lobuchwepheshe bokufakelwa olukwazi ukusisiza sazi ukuthi kwenzekani ebuchosheni bethu
Ukungena emininingwaneni ethe xaxa, njengoba kumenyezelwe ephepheni elishicilelwe yithimba labaphenyi abavela ku- I-Purdue University, eholwa ngu UZhongming Liu, umsizi kaprofesa weBiomedical Engineering and Electrical and Computer Engineering yaleyo yunivesithi, kulo msebenzi bathathe isinqumo sokusebenzisa i- inethiwekhi ye-convolutional neural, uhlobo oluthile lwe-algorithm, ngokwesibonelo, ukusetshenziswa kwalo olubaluleke kakhulu ikakhulukazi lapho kuthuthukiswa ubuhlakani bokufakelwa namuhla esingabuthokozela kuma-smartphones ethu futhi okwazi ukubona ubuso bethu noma izinto ezahlukahlukene.
Kude nohlelo olusetshenziselwe ukuthuthukisa lolu hlelo oluyingqayizivele lobunhloli bokufakelwa, iqiniso ukuthi okushaya umxhwele ngempela ngalo lonke iphrojekthi nocwaningo sebenzisa kanjalo kwenethiwekhi ye-neural. Ngalo mqondo, kufanele kwaziwe ukuthi ikhethe inqubo entsha ekwazi ukusebenzisa lonke uhlelo uku qonda okushiwo umuntu noma kunini. Ngenxa yalokhu, inethiwekhi uqobo iyakwazi ukufunda nokutadisha indlela ubuchopho bomuntu obukwazi ngayo ukucubungula izithombe ezimile noma ezinye izinhlobo zezinto ezibukwayo.
Ngamazwi ka Haiguang wen, umfundi owenza iziqu zobudokotela ePurdue University futhi osebenze ngenkuthalo ekuthuthukiseni le phrojekthi uphawule wathi:
Lesi yisikhathi sokuqala ukuthi lo mbono usetshenziselwe ukubona ukuthi ubuchopho bukwazi kanjani ukucubungula izigcawu zemvelo futhi kuyisinyathelo ekubhekaneni ne-decoding yobuchopho, njengoba abantu bezama ukwenza umqondo wendawo ebonakalayo neyinkimbinkimbi.
Uhlelo olunjalo kumele ekugcineni luqeqeshwe ngokwamukela imizwa evela kubantu abehlukene.
Indlela yokuqeqesha lolu hlelo olusha, kude nokungena emininingwaneni ephathelene ne-algorithm esetshenzisiwe, bekungukufaka abesifazane abathathu ekubukeni okungaphansi kwalokho Izithombe ezi-972 ezahlukahlukene (Sikhuluma ngamahora wokusebenza ayi-11,5). Kuyo yonke le nqubo, ukuqapha ngaso sonke isikhathi umsebenzi okhona ku-cortex ebonakalayo yobuchopho besifazane ngamunye. Le datha kamuva yasetshenziselwa ukuqeqesha uhlelo lobuchwepheshe bokufakelwa.
Ngokuyinhloko umqondo weqembu kwakuwukuhlola ubuchopho besifazane ngamunye njalo ngemizuzwana emibili ukuze kube okwabo. ubuhlakani bokufakelwa obakha kabusha isipiliyoni sokubuka. Ekupheleni kwalolu vivinyo, imiphumela yokuhlolwa ikhombise ukuthi uhlelo selutholakele obekwazi ukucacisa ngempumelelo idatha yokusebenza kwe-magnetic resonance imaging.
Ngokusho kwezitatimende zika- Haiguang wen:
Usebenzisa le ndlela entsha, ungabona ukuthi ubuchopho buqhekeza kanjani indawo ebonakalayo ibe yizicucu, futhi kungenzeka ukuthi uphinde uhlanganise zonke izingcezu ekucindezelweni okuphelele kwendawo yokubuka.