Oko kufuneka ukwazi ukuqonda iteknoloji eguquka kuyo
Unxibelelwano lwee-Neural yimimodeli yekhomputha yeeyunithi ezixhunyiwe okanye iinombolo ezenzelwe ukudlulisa, ukucwangcisa, nokufunda kwiinkcukacha (idatha) ngendlela efanayo nendlela i-neurons (iiseliti ze-nerve) ezisebenza ngayo abantu.
ZeNewural Networks
Kwi-teknoloji, amanxibelelwano e-neural ajwayele ukubizwa njengethungelwano lwe-neural network (ANNs) okanye iminatha ye-neural ukuhlukanisa kwiinkqubo ze-biological neural. Iingcamango eziphambili emva kwe-ANN kukuba ubuchopho bomntu yiyona "khompyutheni" engqongqo kunye nengqiqo ekhoyo. Ngokumisela i-ANN ngokungangoko kunokwenzeka kwisakhiwo kunye nenkqubo yokusetyenziswa kwengcaciso esetyenziswa ngengqondo, abaphandi banethemba lokudala iikhomputha ezithe zafika okanye zagqithisa ubunzima bobunzima babantu. Iifayile ze-Neural ziyinxalenye ebalulekileyo ekuqhubekeni kwangoku kwiinkcukacha zobungcali (AI), ukufunda komatshini (ML), nokufunda okunzulu .
Indlela I-Neural Networks isebenze ngayo: Uthelekiso
Ukuqonda indlela iindlela zokunxibelelana kwee-neural zisebenza ngayo kunye neentlukwano phakathi kweentlobo ezimbini (i-biological and artificial), makhe sisebenzise umzekelo wezakhiwo zeofisi ezili-15 kunye neefowuni zeefowuni kunye neebhodi zokutshintshela ezenza iifowuni kulo lonke isakhiwo, kwindawo nganye, nakwiofisi nganye. I-ofisi nganye kwisakhiwo seofisi yethu ye-15 ibonisa i-neuron (i-node kwi-computer kwinethiwekhi okanye insimbi ye-nerve kwi-biology). Isakhiwo ngokwalo sisakhiwo esineefom yeeofisi ezicwangciswe kwinkqubo ye-15 phantsi (inethiwekhi ye-neural network).
Ukusebenzisa umzekelo kumanethiwekhi e-neural biological, i-switchboard efumana iifowuni inemigca yokudibanisa nayiphi na iofisi kunoma yimuphi umgangatho kwisakhiwo sonke. Ukongezelela, iofisi nganye ineemigca edibanisa nayo yonke i-ofisi kwisakhiwo sonke kwisantya. Khawucinge ukuba ifowuni ifika (ingeniso) kwaye i-switchboard iyithumela kwiofisi kumgangatho wesi-3, leyo idlulisela ngqo kwi-ofisi kumgangatho we-11, leyo iyakusithumela ngqo kwi-ofisi kumgangatho wesi-5. Kwingqondo, i-neuron nganye okanye iisensi ze-nerve (iofisi) inokudibanisa ngokuthe ngqo nakweyiphi enye i-neuron kwinkqubo yayo okanye inethiwekhi ye-neural (isakhiwo). Ulwazi (ifowuni) lunokudluliselwa kunoma iyiphi enye i-neuron (iofisi) ukuba iqhube okanye ifunde oko kuyifunayo de kubekho impendulo okanye isisombululo (ukuphuma).
Xa sisebenzisa lo mzekelo kwi-ANN, iyanzima kakhulu. Umgangatho ngamnye wesakhiwo ufuna i-switchboard yayo, ekwazi ukuxhuma kuphela kwiiofisi kumgangatho ofanayo, kunye nee-boardboards eziphantsi nangaphezulu. Iofisi nganye inokudibanisa ngqo nakwezinye iiofisi kumgangatho ofanayo kunye nebhodi lokutshintshela kuloo mgangatho. Zonke iifowuni kufuneka ziqale ngebhodi lokushiya kwisitezi soku-1 kwaye kufuneka zihanjiswe kumgangatho ngamnye ngamnye ngokulandelana kwamanani ukuya kumgangatho we-15 phambi kokuba umnxeba unokuphela. Masiyibeke phambili ukuze sibone indlela isebenza ngayo.
Cinga ukuba ifowuni ifika (ifakwe) kwi-1 st floorboard switchboard kwaye ithunyelwa kwiofisi kwisigaba sokuqala esingu-1 (node). Le fowuni idluliselwa ngqo phakathi kwezinye iiofisi (iisitifiketi) kwisigaba soku-1 ukuya kulungele ukuthunyelwa kumgangatho olandelayo. Emva koko ifowuni kufuneka ibuyiselwe kwi-1 st floorboard switchboard, leyo iyayidlulisela kumgangatho we-2 phantsi. Ezi nyathelo ezifanayo ziphinda umgangatho omnye ngexesha, kunye nefowuni ithunyelwe kule nkqubo kuwo wonke umgangatho omnye ukuya kwinqanaba eli-15.
Kwi-ANN, ama-nodes (iiofisi) ahlelwe ngamacandelo (phantsi kwesakhiwo). Ulwazi (umnxeba) luhlala lugqityiwe kumgca wokufaka umgangatho (1 ip floor kunye nebhodi lokutshintshela) kwaye kufuneka ithunyelwe kunye nokucutshungulwa ngqimba ngalinye (umgangatho) phambi kokuba liqhubele kwelinye elilandelayo. Uluhlu ngalunye (umgangatho) luqhuba iinkcukacha ezithile malunga naloo fowuni kwaye luthumela umphumo kunye nefowuni kwisalathiso esilandelayo. Xa ifowuni ifinyelela kumgca wokuphuma (umgangatho we-15 kunye nebhodi lokutshintshela), kubandakanya ulwazi lokucwangcisa ukusuka kumacandelo 1-14. Amacandelo (iiofisi) kwisithuba se-15 (umgangatho) usetyenziso lweenkcukacha kunye nokucwangcisa ulwazi kuzo zonke ezinye izalathisi (phantsi) ukuza nempendulo okanye isisombululo (ukuphuma).
I-Neural Networks kunye noFundo lweMishini
Inetha le-Neural yinto enye yethekhnoloji phantsi kodidi lokufunda umatshini. Enyanisweni, ukuqhubela phambili uphando kunye nophuhliso lweenetha ze-neural kuye kwaxhunyaniswa ngokuqinile kwi-ebbs kunye nokuhamba kwe-ML. Iifayile ze-Neural zandisa amandla okucwangciswa kwedatha kwaye zandisa amandla okusebenzisa i-computer, zikhuphula umthamo wolwazi onokucutshungulwa kodwa ukwazi ukwenza imisebenzi enzima.
Imodeli yokuqala yekhompyutheni ebhaliweyo ye-ANN yenziwa ngo-1943 nguWalter Pitts noWarren McCulloch. Inzala yokuqala kunye nophando kwiisetyenziselwano ze-neural kunye nokufunda komatshini ekugqibeleni kwanciphisa kwaye kwakuncinci kancinci ngo-1969, kunye neembalo ezincinane zokuvuselela umdla. Iikhomputha zeso sihlandlo azizange zibe neenkqubo zokukhawuleza ngokwaneleyo okanye ezinkulu ngokwaneleyo ukuqhubela phambili le mimandla eqhubekayo, kwaye ubuninzi beenkcukacha ezifunekayo kwi-ML kunye neenetha ze-neural zazingekho ngelo xesha.
Ukwanda okukhulu kwi-computer ngethuba kunye nokukhula kunye nokwandiswa kwe-intanethi (ngoko ke ukufikelela kwiimali ezinkulu zedatha nge-intanethi) ziye zazisombulula loo mngeni. Iinkonzo ze-Neural kunye ne-ML ngoku zixhobo ekubuchwephesheni esizibonayo kwaye sisebenzise imihla ngemihla, njengokubona kwesobuso , ukucwangciswa komfanekiso kunye nokukhangela, kunye nokuguqulelwa kwexesha langempela-ukubiza nje ezimbalwa.
Imizekelo yeNewural Network kwi-Life Daily
I-ANN yimixholo enzima kwi-teknoloji, nangona kunjalo, kukufanele kuthathe ixesha lokuhlola ngenxa yenani elikhulayo leendlela ezichaphazela ngayo ubomi bethu imihla ngemihla. Nazi ezinye imizekelo embalwa yezindlela zonxibelelwano ze-neural ezisetyenziswa ngamashishini ahlukeneyo ngoku:
- Imali: Iisetyenziso ze-Neural zisetyenziselwa ukuqikelela ixabiso lokutshintshwa kwemali. Ziye zisetyenziswe kwi-teknoloji emva kweenkqubo zorhwebo ngokuzenzekelayo ezisetyenzisiweyo kwimarike yemasheya.
- Amachiza: Ukukhangela imifanekiso yamanetha e-neural kuye kwafaka isandla kwi-teknoloji ekunceda ngakumbi ukuyikhumbuza ngokuchanekileyo kunye nokufumanisa isigaba sokuqala kwaye kunzima ukuchonga iintlobo zeekhensi. Olunye uhlobo olunomdlavuza luyingozi engumhlaza we-melanoma, uhlobo olubi kakhulu kwaye lubulalayo lomhlaza wesikhumba. Ukuchonga i-melanoma kumanqanaba angaphambili, ngaphambi kokusasazeka, kunika izigulane ngale ndlela yomhlaza ithuba elihle lokulibetha.
- Imozulu: Ukukwazi ukufumana utshintsho olummandla olubonisa ukuba isimo sezulu esinobungozi kunye nengozi ngokukhawuleza kwaye ngokunembile kunokwenzeka kubalulekile ekusindiseni ubomi. Iifayile ze-Neural zibandakanyeka ekusebenziseni ixesha langempela le-satellite kunye nemifanekiso ye-radar engabonakali kuphela ukubunjwa kwangaphambili kweziphepheko kunye namaqhwithi, kodwa kwakhona ukufumanisa utshintsho olukhawuleza kwisantya somoya kunye nesalathiso esibonisa i-stormy wind form. Iingqungquthela zizinye zezona ziganeko ezinzulu kunobungozi kakhulu kwaye ziyingozi kwiirekhodi - ngokuphindaphindiweyo ngokukhawuleza, ezonakalisayo, kwaye ezibulalayo kuneziphepho.