Artificial Intelligence (AI) enwetala nnukwu ewu ewu n'afọ ndị na-adịbeghị anya.
Ọ bụrụ na ị bụ onye injinia sọftụwia, ọkà mmụta sayensị kọmputa, ma ọ bụ onye na-anụ ọkụ n'obi sayensị n'ozuzu, mgbe ahụ ị nwere ike ịmasị gị na ngwa ịtụnanya nhazi onyonyo, njirimara ụkpụrụ na nchọpụta ihe nke mpaghara a na-enye.
Mpaghara subfield kachasị mkpa nke AI nke ị nwere ike nụ maka ya bụ mmụta miri emi. Mpaghara a na-elekwasị anya na algọridim dị ike (ntụziaka mmemme kọmputa) emebere ka arụ ọrụ ụbụrụ mmadụ mara dị ka Netiwu Neural.
N'isiokwu a, anyị ga-agafe echiche nke Neural Networks na otu esi ewulite, chịkọta, dabara na nyochaa ụdị ndị a site na iji. Python.
Netiwu Neural
Neural Networks, ma ọ bụ NNs, bụ usoro nke algọridim emebere ka arụ ọrụ ndu nke ụbụrụ mmadụ. Netwọk akwara nwere ọnụ, nke a na-akpọkwa neurons.
A na-akpọ nchịkọta ọnụ ọnụ kwụ ọtọ dị ka akwa. Ụdị ahụ nwere otu ntinye, otu mmepụta, na ọnụ ọgụgụ nke oyi akwa zoro ezo. Nke ọ bụla oyi akwa nwere ọnụ, nke a na-akpọkwa neurons, ebe mgbako na-ewere ọnọdụ.
N'ime eserese na-esonụ, okirikiri na-anọchi anya ọnụ ọnụ na nchịkọta ọnụ nke kwụ ọtọ na-anọchi anya ọkwa. Enwere ọkwa atọ n'ụdị a.
A na-ejikọta ọnụ nke otu oyi akwa na oyi akwa na-esote site na ahịrị nnyefe dị ka a hụrụ n'okpuru.
Nhazi data anyị nwere data akpọrọ. Nke a pụtara na e kenyela ụlọ ọrụ data ọ bụla otu uru aha.
Yabụ maka dataset nhazi ọkwa anụmanụ anyị ga-enwe onyonyo nwamba na nkịta dịka data anyị, ya na 'cat' na 'nkịta' dị ka akara anyị.
Ọ dị mkpa ịmara na ekwesịrị ịgbanwe akara aha ka ọ bụrụ ụkpụrụ ọnụọgụ maka ihe nlereanya anyị iji mee ka uche dị na ha, yabụ akara anụmanụ anyị na-aghọ '0' maka pusi na '1' maka nkịta. A na-agafe ma data ahụ na akara ndị ahụ site na ihe nlereanya.
Learning
A na-enye data n'ụdị otu ihe n'otu oge. A na-agbaji data a n'ime iberibe ma gafere n'ọnụ ọnụ nke ọ bụla nke ụdị ahụ. Ọnụ ụzọ na-arụ ọrụ mgbakọ na mwepụ na akụkụ ndị a.
Ọ dịghị mkpa ka ị mara ọrụ mgbakọ na mwepụ ma ọ bụ mgbako maka nkuzi a, mana ọ dị mkpa inwe echiche zuru oke banyere otu ụdị ndị a si arụ ọrụ. Mgbe usoro nhazi usoro na otu oyi akwa, data na-agafe na oyi akwa na-esote na ihe ndị ọzọ.
Ozugbo emechara, ihe nlereanya anyị na-ebu amụma akara data na oyi akwa mmepụta (dịka ọmụmaatụ, na nsogbu nhazi anụmanụ anyị na-enweta amụma '0' maka pusi).
Ihe nlereanya ahụ gakwara n'ihu iji uru amụma a tụlere na nke ọnụ ahịa akara aha n'ezie.
Ọ bụrụ na ụkpụrụ dakọtara, ihe nlereanya anyị ga-ewere ntinye na-esote ma ọ bụrụ na ụkpụrụ dị iche, ihe nlereanya ahụ ga-agbakọọ ọdịiche dị n'etiti ụkpụrụ abụọ ahụ, nke a na-akpọ ọnwụ, ma gbanwee ọnụ ọnụ ọnụ iji mepụta akara dakọtara oge ọzọ.
Usoro mmụta miri emi
Iji wuo Neural Networks na koodu, anyị kwesịrị ibubata Usoro mmụta miri emi mara dị ka ọba akwụkwọ na-eji anyị Integrated Development Environment (IDE).
Usoro ndị a bụ mkpokọta ọrụ edeburu nke ga-enyere anyị aka na nkuzi a. Anyị ga-eji usoro Keras wuo ihe nlereanya anyị.
Keras bụ ọbá akwụkwọ Python nke na-eji mmụta miri emi na azụ azụ ọgụgụ isi a na-akpọ tensor eruba iji mepụta NN n'ụdị usoro usoro dị mfe na mfe.
Keras na-abịa na ụdị adịla mbụ nke enwere ike iji. Maka nkuzi a, anyị ga-eji Keras mepụta ihe nlereanya nke anyị.
Ị nwere ike mụtakwuo maka usoro mmụta mmụta miri emi site na Ebe nrụọrụ weebụ Keras.
Iwulite netwọkụ akwara ozi ( nkuzi)
Ka anyị gaa n'ihu n'ịrụ netwọkụ Neural site na iji Python.
Nkwupụta Nsogbu
Netwọk Neural bụ ụdị ngwọta maka nsogbu ndị dabeere na AI. Maka nkuzi a, anyị ga-agafe data Pima Indian Diabetes Data, nke dị Ebe a.
ICU Machine Learning achịkọtara dataset a ma nwee ndekọ ahụike nke ndị ọrịa India. Ihe nlereanya anyị ga-ebu amụma ma onye ọrịa nwere mmalite nke ọrịa shuga n'ime afọ 5 ma ọ bụ na ọ nweghị.
Na-ebu Dataset
Nhazi data anyị bụ otu faịlụ CSV akpọrọ 'diabetes.csv' nke enwere ike iji Microsoft Excel mee ya ngwa ngwa.
Tupu ịmepụta ihe nlereanya anyị, anyị kwesịrị ibubata dataset anyị. Iji koodu na-esonụ ị nwere ike ime nke a:
ibubata pandas dika pd
data = pd.read_csv('diabetes.csv')
x = data.dobe ("mpụta")
y = data [“Nsonaazụ”]
Ebe a anyị na-eji Pandas Ọbá akwụkwọ iji nwee ike ijikwa data faịlụ CSV anyị, read_csv() bụ arụnyere arụnyere Pandas nke na-enye anyị ohere ịchekwa ụkpụrụ dị na faịlụ anyị na mgbanwe a na-akpọ 'data'.
Ngbanwe x nwere dataset anyị na-enweghị nsonaazụ (aha) data. Anyị na-enweta nke a site na ọrụ data.drop () na-ewepụ akara maka x, ebe y nwere naanị nsonaazụ (labelụ) data.
Nlereanya Usoro iwu ụlọ
Nzọụkwụ 1: Bubata ọba akwụkwọ
Nke mbụ, anyị kwesịrị ibubata TensorFlow na Keras, yana ụfọdụ paramita achọrọ maka ihe nlereanya anyị. Koodu na-enye anyị ohere ime nke a:
mbubata tensorflow dị ka tf
si keras mbubata tensorflow
si tensorflow.keras.models mbubata Usoro
site na tensorflow.keras.layers mbubata Nrụ ọrụ, oke
site na tensorflow.keras.optimizers na-ebubata Adam
site na tensorflow.keras.metrics na-ebubata categorical_crossentropy
Maka ihe nlereanya anyị, anyị na-ebubata akwa akwa. Ndị a bụ n'ụzọ zuru ezu jikọtara n'ígwé; Ya bụ, ọnụ nke ọ bụla na oyi akwa na-ejikọta ya na ọnụ ọzọ na oyi akwa na-esote.
Anyị na-ebubatakwa ihe bonobo ọrụ dị mkpa maka scaling data ezigara ọnụ. Ndị na-ebuli elu ebubatakwala ya ka o wedata mfu.
Adam bụ onye nrụpụta ama ama nke na-eme ka mgbako ọnụ ọnụ ọnụ anyị melite nke ọma, yana categorical_crossentropy nke bụ ụdị ọrụ mfu (na-agbakọ ọdịiche dị n'etiti ụkpụrụ akara n'ezie na nke amụma) nke anyị ga-eji.
Nzọụkwụ 2: Ịmepụta Ihe Nlereanya Anyị
Ihe nlereanya m na-eke nwere otu ntinye (na nkeji iri na isii), otu zoro ezo (ya na nkeji 16) na otu mmepụta (na nkeji 32). A naghị edozi ọnụọgụ ndị a ma ga-adabere kpamkpam na nsogbu enyere.
Ịtọlite ọnụ ọgụgụ ziri ezi nke nkeji na ọkwa bụ usoro enwere ike imeziwanye oge karịa site na omume. Nkwalite kwekọrọ na ụdị nchacha anyị ga-eme na data anyị tupu anyị agafee ya na ọnụ.
Relu na Softmax bụ ọrụ mgbake ama ama maka ọrụ a.
nlereanya = Usoro ([
Oke( nkeji = 16, ntinye_shape = (1,), ịgbalite = 'relu'),
Oke ( nkeji = 32, ịgbalite = 'relu'),
Oke ( nkeji = 2, ịgbalite = 'softmax')
])
Nke a bụ ihe nchịkọta ihe nlereanya kwesịrị ịdị ka:
Ọzụzụ Model
A ga-azụ ihe nlereanya anyị na nzọụkwụ abụọ, nke mbụ na-achịkọta ihe nlereanya (ịtinye ihe nlereanya ọnụ) na nke ọzọ na-adaba na ihe nlereanya ahụ na dataset nyere.
Enwere ike ime nke a site na iji ọrụ model.compile () na-esote ọrụ model.fit ().
model.compile(njikarịcha = Adam(ọnụego mmụta = 0.0001), ọnwụ = 'binary_crossentropy', metrics = ['nkenke'])
model.fit (x, y, epochs = 30, batch_size = 10)
Ịkọwapụta metrik 'zie ezie' na-enye anyị ohere ịhụ izi ezi nke ihe nlereanya anyị n'oge ọzụzụ.
Ebe ọ bụ na akara anyị dị n'ụdị 1 na 0, anyị ga-eji ọrụ mfu ọnụọgụ abụọ iji gbakọọ ọdịiche dị n'etiti aha n'ezie na nke buru amụma.
A na-ekewakwa dataset ahụ n'ime ogbe iri (batch_size) a ga-agafe n'ụdị ahụ ugboro 10 (epochs). Maka ihe ndekọ data enyere, x ga-abụ data na y ga-abụ akara ndị kwekọrọ na data ahụ.
Nlereanya Nlele Iji amụma
Iji nyochaa ihe nlereanya anyị, anyị na-ebu amụma na data ule site na iji ọrụ amụma ().
amụma = model.predict(x)
Ma nke ahụ bụ ya!
Ugbu a ị ga-enwe nghọta nke ọma nke Ịmụta nke ọma ngwa, Neural Networks, otú ha si arụ ọrụ n'ozuzu na otú e si ewu, ịzụ na nwalee ihe nlereanya na Python code.
Enwere m olileanya na nkuzi a ga-enye gị kickstart imepụta na tinye ụdị mmụta mmụta miri emi nke gị.
Mee ka anyị mara na nkọwa ma ọ bụrụ na isiokwu ahụ bara uru.
Nkume a-aza