UKeras

Njengochwepheshe kuhlelo lwe-Python kanye nohlaka lwe-Keras Deep Learning, ngiyabuqonda ubunkimbinkimbi obuhilelekile ekulayisheni amamodeli, ikakhulukazi uma imodeli yakho isebenzisa umsebenzi wokulahlekelwa ngokwezifiso. Lesi sihloko sikuqondisa ukuthi ungazinqoba kanjani lezi zinselele futhi ulayishe ngempumelelo imodeli yakho ye-Keras ngomsebenzi wokulahlekelwa ngokwezifiso.

I-Keras, i-API yamanethiwekhi e-neural esezingeni eliphezulu, isebenziseka kalula futhi i-modular, ekwazi ukusebenza phezu kwe-TensorFlow noma i-Theano. Kuyaziwa ngobulula nokusebenziseka kalula. Nokho, naphezu kokulula kwayo, ukuqonda imisebenzi ethile njengokulayisha imodeli ngomsebenzi wokulahlekelwa ngokwezifiso kungaba nzima kakhulu.

Funda kabanzi

Izendlalelo zamagama kulo mongo zibhekisela esakhiweni senhlangano esivame ukusetshenziswa ekubhaleni amakhodi, ukwenza amakhodi afundeke kakhudlwana, ahleleke futhi aqondeke kalula. Izendlalelo zamagama zibuye zithuthukise ukusebenza kahle ekusebenzeni kwekhodi ngenxa yesakhiwo sazo esihlelekile esihleliwe. Ukuthola ukuqonda okugcwele kokuthi izendlalelo zamagama zisebenza kanjani kuPython, ake singene emsukeni wenkinga.

Funda kabanzi

Ukwakha imodeli yenethiwekhi ye-neural kuwumkhakha othakazelisayo ekufundeni komshini, ikakhulukazi ePython. Inikeza ububanzi obuningi bokuhlaziya, ukuqagela, kanye nezinqubo zokuthatha izinqumo ezizenzakalelayo. Ngaphambi kokuthi singene ku-nitty-gritty yokwakha inethiwekhi ye-neural yesakhiwo, kubalulekile ukuqonda ukuthi iyini inethiwekhi ye-neural. Empeleni kuwuhlelo lwama-algorithms olusondeza ukwakheka kobuchopho bomuntu, ngaleyo ndlela kwakheka inethiwekhi ye-neural yokwenziwa okuthi, ngenqubo yokuhlaziya ihumushe idatha yezinzwa, icoshe ama-nuances 'angabonakali' ngedatha eluhlaza, njengoba kwenza ubuchopho bethu.

Funda kabanzi

Impela, ake siqale ngesihloko.

Amamodeli okufunda okujulile asephenduke ingxenye ebalulekile yobuchwepheshe enkathini yanamuhla, futhi ama-algorithms ahlukene okwenza ngcono njenge-Adam Optimizer adlala indima ebalulekile ekusebenzeni kwawo. I-Keras, umtapo wezincwadi wePython onamandla futhi okulula ukuwusebenzisa womthombo ovulekile wokuthuthukisa nokuhlola amamodeli okufunda ajulile, usonga imitapo yolwazi yokubala yezinombolo esebenza kahle i-Theano ne-TensorFlow.

Funda kabanzi

I-Keras ilabhulali enamandla futhi ewusizo yokudala amamodeli okufunda ngomshini, ikakhulukazi amamodeli okufunda ajulile. Esinye sezici zayo ukuhlela imodeli yethu ibe umdwebo ukuze uyiqonde kalula futhi uxazulule izinkinga. Kwesinye isikhathi ukusebenzisa i-keras.utils.plot_model kungase kuphonse amaphutha abonisa izimfuneko zesofthiwe ezingekho, ikakhulukazi i-pydot ne-graphviz. Kulindeleke ukuthi uwafake womabili. Noma kunjalo, ngisho nangemva kokuzifaka, usengathola umyalezo wephutha ofanayo. Lokhu kungenxa yezindlela kanye nezilungiselelo zokucushwa ezingasethwanga kahle. Ngalesi sihloko, sizohamba ngenqubo yokuxazulula le nkinga ethile.

Funda kabanzi

Uma ucabanga ukuthi ufuna i-athikili eku-Python igxaza ku-NumPy Arrays, nansi indatshana yakho:

Ngaphambi kokuthi singene kuqala emininingwaneni yezinyathelo ePython, kubalulekile ukuthi siqale siqonde ukuthi ziyini. I-Strides ingumqondo oku-Python othuthukisa kakhulu ukukhohliswa nokuphathwa kwe-array, ikakhulukazi ama-NumPy arrays.. Kusinikeza amandla okuphatha kahle amalungu afanayo ngaphandle kwesidingo sememori eyengeziwe noma izindleko zokubala. Inani le-stride likhomba izinyathelo ezithathwe yiPython lapho unqamula kuhlelo. Manje ake sihlole ukuthi singasisebenzisa kanjani lesi sici esiyingqayizivele ukuxazulula izinkinga.

Funda kabanzi

Emhlabeni wezinhlelo zekhompiyutha, ukuhlangana namaphutha kuyinto evamile. Thatha, isibonelo, i- Iphutha elingukhiye: 'acc' in Python. Leli phutha livamise ukuvela uma ukhiye othile esizama ukuwufinyelela ovela kusichazamazwi ungekho. Ngenhlanhla, i-Python inikeza isixazululo esicacile sokusingatha izinkinga ezinjalo futhi ivimbele ikhodi yakho ukuthi ingashayi. Lokhu kubandakanya ukusebenzisa izinqubo zokuphatha okuhlukile, ukusebenzisa umsebenzi we-get(), noma ukuhlola okhiye ngaphambi kokufinyelela kubo. Ngendlela efanele, leli phutha lingaphathwa ngobuchule.

Funda kabanzi

I-Parametric Rectified Linear Units, noma i-PRELU, iletha ukuguquguquka kwezendlalelo ze-Keras convolution. Njengoba nje imfashini ijwayela ukushintsha izitayela, kanjalo namamodeli akho e-AI angakwazi. Lesi sici sithatha umsebenzi odumile we-Rectified Linear Unit (ReLU) isinyathelo esiqhubekayo ngokuvumela ukuthambekela okunegethivu ukuthi kufundwe kudatha yokufaka, kunokuhlala kugxilile. Ngokwezinto ezibonakalayo, lokhu kusho ukuthi nge-PRELU, amamodeli akho e-AI angakhipha futhi afunde izici ezinhle nezingezinhle kudatha yakho yokufaka, athuthukise ukusebenza kwawo nokusebenza kahle kwawo.

Funda kabanzi