Tese tinoziva nekugona kweneural network yekuziva zvakanyorwa nemaoko. Izvo zvekutanga zvetekinoroji iyi zvave zviripo kwemakore mazhinji, asi ichangopfuura apo kusvetuka musimba rekombuta uye parallel processing kwakaita kuti tekinoroji iyi ive mhinduro inoshanda. Nekudaro, iyi mhinduro inoshanda ingangove iri muchimiro chekombuta yedhijitari inoramba ichichinja mabhiti, senge chero chirongwa. Asi handizvo zvakadaro neneural network yakagadziridzwa nevaongorori kumayunivhesiti eWisconsin, MIT, uye Columbia. Ivo .
Girazi iri rine inclusions yakanyatsowanikwa, senge mabhuru emhepo, kusvibiswa kwegraphene nezvimwe zvinhu. Kana chiedza chichirova girazi, maitiro akaoma esaisai anogadzirwa, zvichiita kuti chiedza chiwedzere kusimba mune imwe kubva munzvimbo gumi. Imwe neimwe yenzvimbo idzi inoenderana nenhamba. Semuenzaniso, pazasi pane mienzaniso miviri inoratidza mafambiro anoita chiedza kana uchiziva nhamba βmbiriβ.
Iine seti yekudzidziswa ye5000 mifananidzo, iyo neural network inokwanisa kuona nemazvo 79% ye1000 yemifananidzo yekuisa. Chikwata chinotenda kuti vanogona kuvandudza mhedzisiro kana vakakwanisa kunzvenga zvipimo zvakakonzerwa nekugadzirwa kwegirazi. Vakatanga neyakaganhurwa dhizaini yechishandiso kuti vawane prototype inoshanda. Tevere, vanoronga kuenderera mberi nekudzidza nzira dzakasiyana dzekuvandudza kunaka kwekuzivikanwa, uku vachiedza kusanyanya kuomesa tekinoroji kuitira kuti igozoshandiswa mukugadzira. Chikwata zvakare chine zvirongwa zvekugadzira XNUMXD neural network mugirazi.
Source: www.habr.com