Artificial Intelligence (AI) yapeza kutchuka kwakukulu m'zaka zaposachedwa.
Ngati ndinu injiniya wamapulogalamu, wasayansi wamakompyuta, kapena wokonda sayansi ya data nthawi zambiri, ndiye kuti mukusangalatsidwa ndi ntchito zodabwitsa za kukonza zithunzi, kuzindikira mawonekedwe ndi kuzindikira kwazinthu zomwe zimaperekedwa ndi gawoli.
Gawo lofunikira kwambiri la AI lomwe mwina mudamvapo ndi Kuphunzira Mwakuya. Gawoli limayang'ana kwambiri ma aligorivimu amphamvu (malangizo a pulogalamu yamakompyuta) opangidwa motengera momwe ubongo wamunthu umagwirira ntchito Ma Neural Networks.
Munkhaniyi, tikambirana za Neural Networks ndi momwe tingamangire, kuphatikiza, kulinganiza ndikuwunika mitundu iyi pogwiritsa ntchito. Python.
Ma Neural Networks
Neural Networks, kapena NNs, ndi mndandanda wa ma algorithms otsatiridwa ndi zochitika zamoyo zaubongo wamunthu. Neural Networks imakhala ndi ma node, omwe amatchedwanso ma neurons.
Kutolere kwa mfundo zoyima kumadziwika kuti zigawo. Chitsanzocho chimakhala ndi kulowetsa kumodzi, kutulutsa kumodzi, ndi zigawo zingapo zobisika. Chigawo chilichonse chimakhala ndi ma node, omwe amatchedwanso ma neurons, komwe kuwerengera kumachitika.
Pachithunzi chotsatirachi, mabwalowo akuyimira ma node ndipo kusonkhanitsa koyima kwa mfundozo kumayimira zigawo. Pali zigawo zitatu mu chitsanzo ichi.
Ma node a gawo limodzi amalumikizidwa ku gawo lotsatira kudzera mu mizere yopatsirana monga momwe tawonera pansipa.
Dongosolo lathu lili ndi zolembedwa. Izi zikutanthauza kuti gulu lililonse la data lapatsidwa mtengo wa dzina.
Chifukwa chake pagulu la nyama tikhala ndi zithunzi za amphaka ndi agalu monga deta yathu, zokhala ndi zilembo za 'mphaka' ndi 'galu'.
Ndikofunika kuzindikira kuti zilembo ziyenera kusinthidwa kukhala manambala kuti chitsanzo chathu chimveke bwino, kotero zilembo zathu zanyama zimakhala '0' pa mphaka ndi '1' pa galu. Deta ndi zolemba zonse zimadutsa mu chitsanzo.
kuphunzira
Deta imaperekedwa ku bungwe limodzi panthawi imodzi. Deta iyi imagawidwa m'magulu ndipo imadutsa mugawo lililonse lachitsanzo. Ma Node amachita masamu pazigawo izi.
Simufunikanso kudziwa masamu kapena kuwerengera kwa phunziroli, koma ndikofunikira kukhala ndi lingaliro lambiri la momwe mitunduyi imagwirira ntchito. Pambuyo pa mawerengedwe angapo mu gawo limodzi, deta imaperekedwa pamtundu wina ndi zina zotero.
Akamaliza, chitsanzo chathu chimaneneratu za zomwe zidzalembedwe (mwachitsanzo, pavuto lamagulu a nyama timalosera '0' kwa mphaka).
Kenako chitsanzocho chimayamba kuyerekeza mtengo womwe wanenedweratuwo ndi wa mtengo wake weniweni.
Ngati zikhalidwe zikugwirizana, chitsanzo chathu chidzatenga chotsatira chotsatira koma ngati zikhalidwe zikusiyana chitsanzocho chidzawerengera kusiyana pakati pa zikhalidwe zonse ziwiri, zomwe zimatchedwa kutayika, ndikusintha mawerengedwe a node kuti apange zilembo zofanana nthawi ina.
Maphunziro Ozama Kwambiri
Kuti mupange Neural Networks mu code, tiyenera kuitanitsa Zofunda zakuya omwe amadziwika kuti malaibulale omwe amagwiritsa ntchito Integrated Development Environment (IDE).
Zomangamangazi ndi mndandanda wa ntchito zomwe zidalembedwa kale zomwe zitithandize mu phunziroli. Tikhala tikugwiritsa ntchito chimango cha Keras kupanga chitsanzo chathu.
Keras ndi laibulale ya Python yomwe imagwiritsa ntchito kuphunzira mozama komanso luntha lochita kupanga lotchedwa Kutuluka kwamatsenga kupanga ma NN mu mawonekedwe osavuta otsatizana mosavuta.
Keras imabweranso ndi mitundu yake yomwe ilipo yomwe ingagwiritsidwenso ntchito. Pa phunziroli, tikhala tikupanga chitsanzo chathu pogwiritsa ntchito Keras.
Mutha kuphunzira zambiri zachikhazikitso cha Deep Learning kuchokera ku Webusaiti ya Keras.
Kupanga Neural Network (Tutorial)
Tiyeni tipitirire kumanga Neural Network pogwiritsa ntchito Python.
Chiwerengero Chavuto
Neural Networks ndi njira yothetsera mavuto a AI. Pa phunziroli tikhala tikupita ku Pima Indians Diabetes Data, yomwe ilipo Pano.
ICU Machine Learning yapanga deta iyi ndipo ili ndi mbiri yachipatala ya odwala aku India. Chitsanzo chathu chiyenera kuneneratu ngati wodwala ali ndi matenda a shuga mkati mwa zaka 5 kapena ayi.
Kutsegula Dataset
Seti yathu ndi fayilo imodzi ya CSV yotchedwa 'diabetes.csv' yomwe imatha kusinthidwa mosavuta pogwiritsa ntchito Microsoft Excel.
Tisanapange chitsanzo chathu, tiyenera kuitanitsa deta yathu. Pogwiritsa ntchito code iyi mutha kuchita izi:
tumizani ma pandas ngati pd
data = pd.read_csv('diabetes.csv')
x = data.drop("Zotsatira")
y = data["Zotsatira"]
Apa tikugwiritsa ntchito Pandas laibulale kuti athe kugwiritsa ntchito mafayilo athu a CSV, read_csv() ndi ntchito yomanga ya Pandas yomwe imatilola kusunga zikhalidwe zomwe zili mufayilo yathu kumitundu yosiyanasiyana yotchedwa 'data'.
The variable x ili ndi deta yathu popanda zotsatira (zolemba) deta. Timakwaniritsa izi ndi ntchito ya data.drop() yomwe imachotsa zilembo za x, pamene y ili ndi zotsatira (label) data yokha.
Kupanga Ma Sequential Model
Khwerero 1: Kulowetsa Malaibulale
Choyamba, tiyenera kuitanitsa TensorFlow ndi Keras, pamodzi ndi magawo ena ofunikira pa chitsanzo chathu. Khodi yotsatirayi imatithandiza kuchita izi:
import tensorflow ngati tf
kuchokera ku tensorflow import keras
kuchokera tensorflow.keras.models import Sequential
kuchokera tensorflow.keras.layers import Activation, Dense
kuchokera tensorflow.keras.optimizers kuitanitsa Adam
kuchokera tensorflow.keras.metrics import categorical_crossentropy
Pachitsanzo chathu tikuitanitsa zigawo zowirira. Awa ndi zigawo zolumikizidwa kwathunthu; mwachitsanzo, mfundo iliyonse mu wosanjikiza imalumikizidwa kwathunthu ndi mfundo ina mugawo lotsatira.
Ifenso tikuitanitsa kunja kutsegulira ntchito yofunikira pakukulitsa deta yotumizidwa ku node. Optimizer zatumizidwanso kuti zichepetse kutaya.
Adam ndiwodziŵika bwino kwambiri omwe amapangitsa kuti mawerengedwe athu a ma node azikhala bwino, komanso categorical_crossentropy chomwe chiri mtundu wa ntchito yotayika (imawerengera kusiyana pakati pa zizindikiro zenizeni ndi zonenedweratu) zomwe tidzakhala tikugwiritsa ntchito.
Gawo 2: Kupanga Chitsanzo Chathu
Mtundu womwe ndimapanga uli ndi cholowetsa chimodzi (chokhala ndi mayunitsi 16), chobisika (chokhala ndi mayunitsi 32) ndi chotulutsa chimodzi (chokhala ndi mayunitsi awiri). Ziwerengerozi sizinakhazikitsidwe ndipo zidzadalira kwambiri vuto lomwe laperekedwa.
Kukhazikitsa chiwerengero choyenera cha mayunitsi ndi zigawo ndi njira yomwe ingawongoleredwe nthawi yowonjezera pochita. Kutsegula kumafanana ndi mtundu wa makulitsidwe omwe tidzakhala tikuchita pa data yathu tisanadutse mu node.
Relu ndi Softmax ndi ntchito zodziwika bwino zoyambitsa ntchitoyi.
chitsanzo = Sequential ([
Ndili (mayunitsi = 16, input_shape = (1,), activation = 'relu'),
wandiweyani (mayunitsi = 32, activation = 'relu'),
Wandiweyani(mayunitsi = 2, activation = 'softmax')
])
Izi ndi zomwe chidule cha chitsanzocho chiyenera kuwoneka:
Kuphunzitsa Chitsanzo
Chitsanzo chathu chidzaphunzitsidwa m'masitepe awiri, choyamba ndikulemba chitsanzo (kuyika chitsanzo pamodzi) ndipo chotsatiracho chidzakhala choyenerera pamtundu wina wa data.
Izi zitha kuchitika pogwiritsa ntchito ntchito ya model.compile() yotsatiridwa ndi ntchito ya model.fit().
model.compile(optimizer = Adam(learning_rate = 0.0001), loss = 'binary_crossentropy', metrics = ['accuracy'])
model.fit(x, y, epochs = 30, batch_size = 10)
Kufotokozera za 'kulondola' kumatilola kuwona kulondola kwachitsanzo chathu panthawi yamaphunziro.
Popeza zilembo zathu zili ngati 1's ndi 0's, tikhala tikugwiritsa ntchito njira yotayika ya binary kuti tiwerengere kusiyana pakati pa zilembo zenizeni ndi zonenedweratu.
Zosungirazo zikugawidwanso m'magulu a 10 (batch_size) ndipo zidutsamo nthawi 30 (epochs). Pa data yomwe yaperekedwa, x ingakhale datayo ndipo y ingakhale zilembo zogwirizana ndi datayo.
Chitsanzo Choyesera Pogwiritsa Ntchito Zolosera
Kuti tiwunikire chitsanzo chathu, timalosera pamayeso pogwiritsa ntchito predict() function.
kulosera = model.predict(x)
Ndipo ndizo zonse!
Tsopano muyenera kumvetsetsa bwino za Kuphunzira Kwambiri kugwiritsa ntchito, Neural Networks, momwe amagwirira ntchito nthawi zonse komanso momwe angapangire, kuphunzitsa ndi kuyesa chitsanzo mu code ya Python.
Ndikukhulupirira kuti phunziroli likupatsani chiyambi chopangira ndikugwiritsa ntchito zitsanzo zanu za Deep Learning.
Tiuzeni mu ndemanga ngati nkhaniyi inali yothandiza.
Siyani Mumakonda