Artificial Intelligence (AI) tau txais txiaj ntsig tseem ceeb hauv xyoo tas los no.
Yog hais tias koj yog ib tug software engineer, computer scientist, los yog cov ntaub ntawv science enthusiast nyob rau hauv dav dav, ces tej zaum koj yuav intrigued los ntawm cov amazing daim ntaub ntawv ntawm cov duab ua, qauv paub thiab cov khoom nrhiav tau los ntawm daim teb no.
Qhov tseem ceeb tshaj plaws subfield ntawm AI uas koj yuav hnov txog yog Deep Learning. Daim teb no tsom rau cov algorithms muaj zog (cov lus qhia hauv computer program) ua qauv tom qab tib neeg lub hlwb ua haujlwm hu ua Neural Tes hauj lwm.
Hauv tsab xov xwm no, peb yuav mus dhau lub tswvyim ntawm Neural Networks thiab yuav ua li cas los tsim, sau, haum thiab ntsuas cov qauv siv. Nab hab sej.
Neural Tes hauj lwm
Neural Networks, lossis NNs, yog cov txheej txheem ntawm cov qauv ua qauv tom qab kev ua haujlwm lom neeg ntawm tib neeg lub hlwb. Neural Networks muaj cov nodes, tseem hu ua neurons.
Ib pawg ntawm cov kab ntsug yog hu ua txheej. Cov qauv muaj ib qho kev tawm tswv yim, ib qho kev tso tawm, thiab ntau cov txheej txheem zais. Txhua txheej muaj cov nodes, tseem hu ua neurons, qhov chaw ntawm kev suav.
Hauv daim duab hauv qab no, cov voj voog sawv cev rau cov nodes thiab cov kab ntsug ntawm cov nodes sawv cev rau cov khaubncaws sab nraud povtseg. Muaj peb txheej hauv cov qauv no.
Cov nodes ntawm ib txheej yog txuas mus rau txheej tom ntej los ntawm kev sib kis kab raws li pom hauv qab no.
Peb dataset muaj cov ntaub ntawv sau npe. Qhov no txhais tau hais tias txhua qhov chaw cov ntaub ntawv tau muab rau qee lub npe tus nqi.
Yog li rau cov ntaub ntawv faib tsiaj ntawv peb yuav muaj cov duab ntawm miv thiab dev raws li peb cov ntaub ntawv, nrog 'miv' thiab 'dub' ua peb cov ntawv.
Nws yog ib qho tseem ceeb uas yuav tsum nco ntsoov tias cov ntawv cim yuav tsum tau hloov mus rau qhov muaj nuj nqis rau peb cov qauv kom paub txog lawv, yog li peb cov tsiaj ntawv ua '0' rau miv thiab '1' rau dev. Ob qho tib si cov ntaub ntawv thiab cov ntawv sau tau dhau los ntawm tus qauv.
kawm
Cov ntaub ntawv yog pub rau tus qauv ib qhov chaw ntawm ib lub sij hawm. Cov ntaub ntawv no tau tawg mus rau hauv chunks thiab dhau los ntawm txhua qhov ntawm tus qauv. Nodes ua cov haujlwm lej ntawm cov chunks.
Koj tsis tas yuav paub txog kev ua lej lossis kev suav rau qhov kev qhia no, tab sis nws tseem ceeb heev kom muaj lub tswv yim dav dav ntawm cov qauv no ua haujlwm li cas. Tom qab ib tug series ntawm kev xam nyob rau hauv ib txheej, cov ntaub ntawv yog kis mus rau lwm lub txheej thiab thiaj li nyob rau.
Thaum ua tiav, peb tus qauv kwv yees cov ntaub ntawv daim ntawv lo ntawm cov khoom tso tawm (piv txwv li, hauv qhov teeb meem kev faib tsiaj peb tau txais kev kwv yees '0' rau miv).
Cov qauv tom qab ntawd pib los sib piv tus nqi kwv yees no nrog rau tus nqi ntawm daim ntawv lo tiag.
Yog tias qhov tseem ceeb sib phim, peb tus qauv yuav siv cov tswv yim tom ntej tab sis yog tias qhov tseem ceeb sib txawv tus qauv yuav xam qhov sib txawv ntawm ob qho txiaj ntsig, hu ua poob, thiab kho cov kev suav cov lej los tsim cov ntawv sib txuam rau lwm zaus.
Deep Learning Frameworks
Txhawm rau tsim Neural Networks hauv code, peb yuav tsum tau import Deep Learning moj khaum lub npe hu ua cov tsev qiv ntawv siv peb qhov Kev Txhim Kho Ib puag ncig (IDE).
Cov txheej txheem no yog cov sau ua ntej sau ua haujlwm uas yuav pab peb hauv qhov kev qhia no. Peb yuav siv Keras lub moj khaum los tsim peb tus qauv.
Keras yog lub tsev qiv ntawv Python uas siv qhov kev kawm sib sib zog nqus thiab kev txawj ntse sab nraud hu ua tensor ntws los tsim NNs nyob rau hauv daim ntawv ntawm cov qauv sib txuas yooj yim nrog yooj yim.
Keras kuj los nrog nws tus kheej preexisting qauv uas yuav siv tau thiab. Rau qhov kev qhia no, peb yuav tsim peb tus kheej qauv siv Keras.
Koj tuaj yeem kawm paub ntau ntxiv txog qhov Deep Learning moj khaum ntawm no Keras lub website.
Tsim kom muaj Neural Network (Kev Qhia)
Cia peb mus rau kev tsim Neural Network siv Python.
Cov Lus Qhia Teeb Meem
Neural Networks yog hom kev daws teeb meem rau AI-raws li cov teeb meem. Rau qhov kev qhia no peb yuav mus dhau Pima Indians Diabetes Data, uas muaj no.
ICU Machine Learning tau sau cov ntaub ntawv no thiab muaj cov ntaub ntawv kho mob ntawm cov neeg mob Indian. Peb tus qauv yuav tsum kwv yees seb tus neeg mob puas muaj qhov pib mob ntshav qab zib hauv 5 xyoos lossis tsis tau.
Loading Dataset
Peb cov ntaub ntawv yog ib daim ntawv CSV hu ua 'diabetes.csv' uas tuaj yeem siv tau yooj yim siv Microsoft Excel.
Ua ntej tsim peb tus qauv, peb yuav tsum tau import peb dataset. Siv cov cai hauv qab no koj tuaj yeem ua qhov no:
pandas import li pd
data = pd.read_csv('diabetes.csv')
x = data.drop("Tsim")
y = data[“Tau”]
Ntawm no peb siv lub pandas tsev qiv ntawv kom muaj peev xwm tswj tau peb cov ntaub ntawv CSV, read_csv() yog qhov ua haujlwm ntawm Pandas uas tso cai rau peb khaws cov txiaj ntsig hauv peb cov ntaub ntawv mus rau qhov sib txawv hu ua 'cov ntaub ntawv'.
Qhov sib txawv x muaj peb dataset yam tsis muaj qhov tshwm sim (labels) cov ntaub ntawv. Peb ua tiav qhov no nrog data.drop() muaj nuj nqi uas tshem tawm cov ntawv rau x, thaum y tsuas muaj qhov tshwm sim (label) cov ntaub ntawv.
Kev tsim qauv ua ntu zus
Kauj Ruam 1: Ntshuam Cov Tsev Qiv Ntawv
Ua ntej, peb yuav tsum tau import TensorFlow thiab Keras, nrog rau qee yam tsis xav tau rau peb cov qauv. Cov cai hauv qab no tso cai rau peb ua qhov no:
import tensorflow li tf
los ntawm tensorflow import keras
los ntawm tensorflow.keras.models import Sequential
los ntawm tensorflow.keras.layers import Activation, Dense
los ntawm tensorflow.keras.optimizers import Adas
los ntawm tensorflow.keras.metrics import categorical_crossentropy
Rau peb cov qauv peb yog importing tuab txheej. Cov no yog tag nrho cov khaubncaws sab nraud povtseg; piv txwv li, txhua qhov ntawm ib txheej yog tag nrho txuas nrog lwm tus ntawm cov txheej tom ntej.
Peb kuj tseem import ib ua kom muaj nuj nqi xav tau rau scaling cov ntaub ntawv xa mus rau nodes. Optimizers kuj tau raug import los txo qhov poob.
Adas yog lub npe hu ua optimizer uas ua rau peb cov qauv hloov kho ntawm cov lej suav tau zoo dua, nrog rau categorical_crossentropy uas yog hom kev poob haujlwm (xws li qhov sib txawv ntawm qhov tseeb thiab kwv yees qhov tseem ceeb ntawm daim ntawv lo) uas peb yuav siv.
Kauj Ruam 2: Tsim Peb Tus Qauv
Cov qauv kuv tsim muaj ib qho kev tawm tswv yim (nrog 16 units), ib qho zais (nrog 32 units) thiab ib qho kev tso tawm (nrog 2 units) txheej. Cov lej no tsis raug kho thiab yuav nyob ntawm qhov teeb meem muab.
Kev teeb tsa tus naj npawb ntawm cov chav nyob thiab cov khaubncaws sab nraud povtseg yog cov txheej txheem uas tuaj yeem txhim kho ua haujlwm dhau sijhawm los ntawm kev xyaum. Kev ua kom raug sib raug rau hom kev ntsuas peb yuav ua haujlwm ntawm peb cov ntaub ntawv ua ntej dhau los ntawm ib qho ntawm qhov.
Relu thiab Softmax yog lub npe nrov ua haujlwm ua haujlwm rau txoj haujlwm no.
model = Sequential([
Dense(units = 16, input_shape = (1,), activation = 'relu'),
Dense(units = 32, activation = 'relu'),
Dense(units = 2, activation = 'softmax')
])
Ntawm no yog cov ntsiab lus ntawm tus qauv yuav tsum zoo li:
Kev cob qhia tus qauv
Peb cov qauv yuav raug cob qhia nyob rau hauv ob kauj ruam, thawj zaug muab cov qauv (tsim tus qauv ua ke) thiab tom ntej no yog haum tus qauv ntawm ib daim ntawv teev npe.
Qhov no tuaj yeem ua tiav siv model.compile() muaj nuj nqi ua raws li model.fit() muaj nuj nqi.
model.compile(optimizer = Adas(learning_rate = 0.0001), poob = 'binary_crossentropy', metrics = ['accuracy'])
model.fit(x, y, epochs = 30, batch_size = 10)
Kev qhia txog qhov 'qhov tseeb' metric tso cai rau peb los soj ntsuam qhov tseeb ntawm peb tus qauv thaum kev cob qhia.
Txij li thaum peb daim ntawv lo yog nyob rau hauv daim ntawv ntawm 1's thiab 0's, peb yuav siv lub binary poob muaj nuj nqi los xam qhov sib txawv ntawm cov ntawv tseeb thiab kwv yees.
Cov dataset tseem raug faib ua pawg ntawm 10 (batch_size) thiab yuav dhau los ntawm tus qauv 30 zaug (epochs). Rau cov ntaub ntawv muab, x yuav yog cov ntaub ntawv thiab y yuav yog cov ntawv sau sib raug rau cov ntaub ntawv.
Kev sim qauv siv kev kwv yees
Txhawm rau ntsuas peb tus qauv, peb ua kev kwv yees ntawm cov ntaub ntawv xeem siv qhov kev kwv yees () ua haujlwm.
twv = model.predict(x)
Thiab tus ntawd yog nws!
Tam sim no koj yuav tsum muaj kev nkag siab zoo ntawm cov Kev kawm tob daim ntawv thov, Neural Networks, lawv ua haujlwm li cas thiab tsim, cob qhia thiab sim tus qauv hauv Python code.
Kuv vam tias qhov kev qhia no muab rau koj lub kickstart los tsim thiab siv koj tus kheej cov qauv kev kawm tob.
Qhia rau peb paub hauv cov lus yog tias tsab xov xwm muaj txiaj ntsig.
Sau ntawv cia Ncua