Shabakadaha neural Artificial

shabakadaha neural Artificial - waa kuwa ka sameysan yihiin unugyo gaar ah - neurons. Waxay yihiin dad lagu daydo xisaabeed ee neurons noolaha, ie, unugyada uu ka samaysan yahay habka dareemayaasha jirka aadanaha.

Waayo, markii ugu horeysay ee aan ka hadlayno shabakadaha neural ee 1943, oo ka dib markii la alifay ee Perceptron Rosenblatt yimid Zaman oo dahabka ahayd, iyo shabakadaha aad caan u noqday. Si kastaba ha ahaatee, ka dib markii la daabaco ee Minsk ee 1969, kaas oo cilmiga ah ayaa cadeeyay karti ee Perceptron, xaaladaha qaarkood, danta waaxdan soo maratay dhacay. Laakiin sheekada kuma dhammaanayo shabakadaha aan dabiici ahayn. . In 1985, J. Hopfield bandhigay waxbarashadooda iyo caddeeyeen in shabakadda neural - qalab weyn oo loogu mashiinka barashada.

Waxaa laga soo amaahday biology dhowr fikradaha iyo mabaadii'da. Neuron - nooc ka mid ah ku biiro oo uu helo, ka dibna gudbiya digir ah (calaamadaha). Haddii neuron helo dardar ku filan oo awood badan, waxaa la rumeysan yahay in ay la shaqeysiin iyo gudbiya karsado ee neurons xidhiidh la leh haray. Neuron isla kaas oo aan shaqeysiin, waxa wali nasan, waxa aanu u gudbin garaaca wadnaha. Neuron ka kooban yahay qaybaha ugu muhiimsan ee dhowr: synapses isku xidha neurons in kasta oo kale oo aad hesho karsado, axon, taasoo baahin falalkooda hawsha iyo dendrites, taas oo uu helo signalada oo laga helay ilo kala duwan. Marka neuron helo Toonsaadaan kor marinka gaar ah, waxa isla markiiba signal a in neuron soo socda.

Habka xisaabta waa wax yar oo kala duwan. Login model xisaabeed ee neuron ah - waa dulinka ah, oo ka kooban tiro badan oo ah qaybaha. Mid kasta oo ka qayb ah - waa mid ka mid ah la karsado, kuwaas oo helay by neuron ah. wax soo saarka ee model ah waa tiro hal. Taasi waa, in dulinka model talooyin la beddelay scalar ah, ka dib wareejiyay neurons kale.

shabakadaha neural loo tababaray kartaa laba siyaabood: qaba oo aan macallin. nidaamka waxbarashada waxay ka kooban tahay dhowr tallaabo. First, on network waa talooyin ka heshaa ka baxsan. Markaas, iyadoo la raacayo xeerarka kala duwan ku soo koobin free of network neural, ka dibna shabakadda jawaab ku meersan talooyin horay duwan. Geedi socodka waa in lagu celiyo ilaa iyo inta network ma xal u helidda dhibaatada. geynta The barashada macalinka waa in inta lagu guda jiro tababarka network ayaa durbaba leh jawaabta saxda ah. Habkani ayaa si guul leh loo isticmaalo codsiyada badan, laakiin waxaa inta badan dhaleeceeyay xaqiiqda ah in ay tahay bayoloji ku sugun tahay. shabakadaha neural waxaa loo tababaray macalinka aan kiiska halkaas oo gashiga Ogtihiin. Iyada oo ku saleysan iyaga on, network si tartiib tartiib ah u barto in la siiyo ku soo saarka qiimaha ugu wanaagsan.

Codsiga shabakadaha neural waa run ahaantii kala duwan. Waxaa inta badan loo isticmaalaa in lagu qalabayn aqoonsi, saadaalinta, abuurka kala duwan ee nidaamka khabiir, qiyaas of functionals. Iyada oo network noocan oo kale ah qaban karto aqoonsi dhawaaqa ama calaamadaha indhaha in la saadaaliyo tilmaamayaasha sarrifka abuuro nidaam awood is-barasho, taas karaa, tusaale ahaan, in synthesize hadalka qoraalka ah la siiyo ama meesha baabuurta la dhigto. shabakadaha neural ee West waxaa loo isticmaalaa firfircoon, nasiib daro, shirkadaha guryaha aan weli soo qaatay habkan.

Inkastoo faa'iidooyinka ANN on xisaabinta caadiga ah meelaha qaarkood, shabakadaha neural jira - ma xal ku haboon. Tan iyo markii ay yihiin awood waxbarashada, waxa laga yaabaa inay qalad ah. Intaa waxaa dheer, in aad si sax ah ma ballan qaadi karno in network horumartay neural waa fiicnayn. sameeysa waa in la fahmo nooca dhibaatada la hagaajiyaa, waxay leeyihiin macluumaad badan oo qeexaya dhibaatada, si ay u helaan xogta baaritaanka iyo network tababarka, si ay u doortaan hab sax ah oo tababar ah, shaqo kala iibsiga iyo hawlaha jilbis.