Okudingayo ukwazi ukuqonda ubuchwepheshe obuseduze kwakho
Amanethiwekhi e-Neural amamodeli wekhompiyutha wezingxenye ezixhunyiwe noma ama-node ayenzelwe ukudlulisa, ukucubungula, nokufunda kusuka kolwazi (idatha) ngendlela efanayo nendlela i-neurons (amangqamuzana omzimba wesibindi) asebenza ngayo kubantu.
I-Neural Networks yokufakelwa
Ngobuchwepheshe, amanethiwekhi we-neural avame ukubhekwa njengama-neural amanethiwekhi (ANNs) noma amanetha we-neural ukuze ahlukanise namanethiwekhi we-neural we-biological awalandelwa ngemuva. Umqondo oyinhloko we-ANN wukuthi ubuchopho bomuntu yi-"computer" eyinkimbinkimbi kunazo zonke futhi ehlakaniphile ekhona. Ngokwenza isibonelo se-ANN ngokuseduze ngangokunokwenzeka kwisakhiwo kanye nesistimu yokucubungula ulwazi esetshenziswa ubuchopho, abacwaningi babe nethemba lokudala amakhompiyutha afike noma aphumelele ukuhlakanipha kwabantu. Amanetha e-Neural ayingxenye ebalulekile ekuthuthukiseni kwamanje ku -intelligence yokufakelwa (AI), ukufunda imishini (ML), nokufunda okujulile .
Indlela I-Neural Networks Esebenza Ngayo: Ukuqhathaniswa
Ukuze uqonde ukuthi ama-neural amanethiwekhi asebenza kanjani futhi umehluko phakathi kwalezi zinhlobo ezimbili (eziphilayo kanye nokufakelwa), ake sisebenzise isibonelo sezakhiwo zehhovisi lezitezi ezingu-15 kanye nemigqa yocingo kanye namabhodi wokushintshela ohambisa amakholi kulo lonke isakhiwo, ngamanye amazwe, namahhovisi ngabanye. Ihhovisi ngalinye ehhovisi lethu lezakhiwo ezingu-15 lisho i-neuron (i-node ekwenzeni i-computer noma i-nerve cell in biology). Isakhiwo ngokwayo isakhiwo esinezigcawu zamahhovisi ahlelwe ohlelweni lwezinsika ezingu-15 (inethiwekhi ye-neural).
Ukusebenzisa lesi sibonelo kumanethiwekhi omzimba we-neural, i-switchboard eyamukela izingcingo inemigqa yokuxhuma kunoma yiliphi ihhovisi kunoma yisiphi isakhiwo kuso sonke isakhiwo. Ukwengeza, ihhovisi ngalinye linemigqa elixhuma kuyo yonke ihhovisi kulo lonke isakhiwo kunoma iyiphi indawo. Cabanga ukuthi ucingo lufika (okokufaka) futhi i-switchboard iyithumele ehhovisi kumgangatho we-3, okuyidlulisela ngqo ehhovisi kwisitezi esingu-11, okuyi-ke ngokuyihambisa ngqo ehhovisi esiteji sesihlanu. Esikhathini sobuchopho, i-neuron ngayinye noma i-cell nerve (ihhovisi) ingaxhuma ngokuqondile kunoma iyiphi enye i-neuron ohlelweni layo noma inethiwekhi ye-neural (isakhiwo). Ulwazi (ucingo) lungadluliselwa kunoma iyiphi enye i-neuron (ihhovisi) ukucubungula noma ukufunda okudingekayo kuze kube yilapho kukhona impendulo noma isinqumo (okukhishwe).
Uma sisebenzisa lesi sibonelo kuma-ANN, sithola kancane kunzima kakhulu. Igumbi ngalinye lesakhiwo lidinga i-switchboard yayo, engakwazi ukuxhuma kuphela kumahhovisi aphansi, kanye namabhodibhodi ezansi phansi nangaphansi kwalo. Ihhovisi ngalinye lingakwazi ukuxhuma ngqo kumahhovisi e-floor efanayo kanye ne-switchboard yaleso sigaba. Wonke amakholi amasha kufanele aqale ngebhodi lokushintshela esitezi sokuqala futhi kufanele adluliselwe kwisitela ngasinye ngabanye ngokulandelana kwamanani kuze kube sezingeni lesishiyagalolunye ngaphambi kokuba ucingo luphele. Ake siyihambise ukuze sibone ukuthi isebenza kanjani.
Cabanga ukuthi ikholi ingena (ukufaka) ku-1 st floorboardboard futhi ithunyelwa ehhovisi kwisitezi esingu- 1 (i-node). Ikholi idluliselwa ngqo phakathi kwamanye amahhovisi (ama-node) esiteji esingu-1 kuze kube yilapho isikulungele ukuthunyelwa esiteji esilandelayo. Khona-ke ucingo kumele lubuyiselwe emuva ku-1 floor floorboard, bese lidlulisela ku-switchboard yesithathu. Lezi zinyathelo ezifanayo ziphinda phansi eyodwa ngesikhathi esisodwa, lapho ucingo luhanjiswa ngale nqubo kuzo zonke izitezi kuze kube sezingeni lesishiyagalolunye.
E-ANN, ama-nodes (amahhovisi) ahlelwe ngezigaba (phansi kwesakhiwo). Ukwaziswa (ucingo) njalo kungena kumgca wokufakelwa (1 st floor kanye ne-switchboard) futhi kufanele ithunyelwe futhi kusetshenziswe ngqimba ngalinye (phansi) ngaphambi kokuthi lihambise elilandelayo. Ingqimba ngayinye (phansi) ihlela imininingwane ethile mayelana nale kholi iphinde ithumele umphumela kanye nekholi kuya kwendlalelo elandelayo. Uma ucingo lufinyelela esendleleni yokukhipha (isitezi esingu-15 th kanye ne-switchboard yayo), kufaka phakathi ulwazi lokucubungula kusuka kuzendlalelo 1-14. Ama-nodes (amahhovisi) e-15 th ungqimba (phansi) asebenzise imininingwane yokufaka nokucubungula kuzo zonke ezinye izendlalelo (phansi) ukuza nempendulo noma isinqumo (ukukhishwa).
I-Neural Networks ne-Learning Machine
Amanetha e-Neural wuhlobo olulodwa lobuchwepheshe ngaphansi kwesigaba sokufunda somshini. Eqinisweni, ukuthuthukiswa ocwaningweni nokuthuthukiswa kwamanetha e-neural kuye kwaxhunyaniswa ngokuqinile kwi-ebbs nokugeleza kokuthuthukiswa ku-ML. Amanetha e-Neural athuthukisa amandla okucubungula idatha futhi akhuthaze amandla e-computing ye-ML, okwandisa ivolumu yedatha engacubungulwa kodwa futhi nekhono lokwenza imisebenzi eyinkimbinkimbi.
I-model yokuqala yekhompyutha ye-ANN yasungulwa ngo-1943 nguWalter Pitts noWarren McCulloch. Intshisekelo yokuqala nokucwaninga kumanethiwekhi omzimba we-neural nokufunda komshini kwagcina kwehla futhi kwasungulwa kancane ngo-1969, nje kuphela ukuqhuma okuncane kwe-interest. Amakhompiyutha ngaleso sikhathi ayengenawo ama-processor akwanele noma amancane okwandisa lezi zindawo ngokuqhubekayo, futhi inani elikhulu ledatha elidingekayo ku-ML namanetha e-neural ayengatholakali ngaleso sikhathi.
Ukunyuka okukhulu kwamakhompiyutha ngokuhamba kwesikhathi kanye nokukhula nokwandiswa kwe-intanethi (futhi ngaleyo ndlela ukufinyelela kwamanani amakhulu edatha nge-intanethi) baye baxazulula lezo zinselelo zakuqala. Amanetha e-Neural noML manje asebambe iqhaza kwezobuchwepheshe esizibonayo futhi sisebenzise nsuku zonke, njengokubona kombuso , ukucubungula izithombe nokucinga, nokuhumusha ulimi lwesikhathi sangempela - ukubiza nje ezimbalwa.
Izibonelo ze-Neural Network ku-Daily Life Life
I-ANN yisifundo esiyinkimbinkimbi ngaphakathi kobuchwepheshe, noma kunjalo, kufanelekile ukuthatha isikhathi esithile sokuhlola ngenxa yenani elikhulayo lezindlela elithinta impilo yethu nsuku zonke. Nazi ezinye izibonelo ezimbalwa zezindlela zokuxhumana ze-neural okwamanje ezisetshenziswa izimboni ezahlukene:
- Ezezimali: Amanetha aseNesural asetshenziselwa ukubikezela izinga lokushintshwa kwezimali. Zibuye zisetshenziswe kwezobuchwepheshe ngemuva kwezinqubo zokuhweba okuzenzakalelayo ezisetshenziselwa imakethe yamasheya.
- Imithi: Amandla okucubungula imidwebo yamanetha e-neural abambe iqhaza kwezobuchwepheshe obasiza ngokucacile ukuyihlola futhi ukuthole isigaba sokuqala futhi kunzima ukubona izinhlobo zegciwane. Olunye uhlobo lomdlavuza luyi-melanoma engavamile, uhlobo olubi nakakhulu lokubulala umdlavuza wesikhumba. Ukubona i-melanoma ezinyathelweni zangaphambili, ngaphambi kokuba isakazeke, inikeza iziguli ngale ndlela uhlobo lomdlavuza ithuba elihle lokulishaya.
- Isimo sezulu: Ikhono lokubona izinguquko zomoya ezitshengisa umcimbi wesimo sezulu esibi futhi esiyingozi ngokushesha futhi ngokunembile ngangokunokwenzeka kubalulekile ekusindiseni izimpilo. Amanetha aseNeural ahileleke ekusebenzeni kwangempela kwesikhathi esithathwe yi-satellites nemifanekiso e-radar engatholakali kuphela ukwakheka kwesikhashana kweziphepho nezinkanyezi, kodwa futhi ukuthola ukushintshwa okungazelelwe emshinini womoya nokuqondisa okubonisa isiphepho esinamandla. Izinkanyamba ziyizenzakalo zezulu eziqine kakhulu nezingozi kunazo zonke ekurekhodini - ngokuphindaphindiwe ngokuzumayo, ezonakalisa futhi ezibulalayo kuneziphepho.