Isingeniso
Ukwenza imodeli nokubikezela okususelwe kuhlu lwamanani okokufaka kuyindlela evamile ekuhlaziyeni idatha nokufunda komshini, kusetshenziswa i-Python njengolimi oluyinhloko lohlelo. Kubandakanya ukulungiselela idatha, ukwakha imodeli ebikezelayo, kanye nokuhlela ikhodi ukuze uthole imiphumela efanele. Ingxenye eyisisekelo enegalelo kule misebenzi i-Tensor, uhlaka lwedatha olusetshenziswa ezinhlelweni ze-Python, ikakhulukazi ngaphakathi kwemitapo yolwazi i-TensorFlow ne-PyTorch. I-tensor ngayinye iqukethe isethi yamanani akudala abunjwe aba uchungechunge lwanoma iyiphi inombolo yobukhulu.
Inkinga Nesixazululo Sayo
Lapho siphatha amasethi edatha ku-ecosystem yokufunda komshini, sivamise ukuhlangana nezinkinga ezihlobene nokuma kwedatha nosayizi. Isibonelo, lapho ucubungula idatha yokokufaka yamanethiwekhi e-neural, ukuhluka kobukhulu bedatha kungadala izinkinga. Isixazululo salezi zinkinga ngokuvamile sitholakala ekusebenzeni kwe-tensor. I-Python, enemitapo yolwazi enamandla njenge-NumPy, i-TensorFlow, ne-PyTorch, inikeza izindlela ezisebenzayo zokuphatha, ukukhohlisa, nokwenza izibikezelo usebenzisa ama-tensor.
[b]Ama-Tensor asebenzayo we-Predictive Modelling[/b]
Ake siqalise i-TensorFlow futhi sichaze inkinga. Cabanga ukuthi sinohlu lwama-tensor (amalungu afanayo anezinhlangothi eziningi ezinohlobo lwedatha efanayo), futhi sifuna ukudala imodeli ebikezelayo sisebenzisa lolu hlu.
Isinyathelo esiyinhloko sokudala imodeli yokubikezela ukulayisha nokulungiselela idatha. Esimeni sethu, sisebenzisa ama-tensor, angase adinge ukubunjwa kabusha kanye nokwejwayelekile ngokwezidingo zedathasethi.
import tensorflow as tf # Assume we have the following list of tensors tensor_list = [tf.constant([2]), tf.constant([5]), tf.constant([10])] # combine these tensors into one big tensor input_tensor = tf.concat(tensor_list, 0) # normalize the tensor input_tensor = tf.nn.l2_normalize(input_tensor, 0)
Lapha, singenise i-TensorFlow, sakha uhlu lwama-tensor, sahlanganisa lawa ma-tensor, futhi sakwenza okokufaka kube okujwayelekile.
Ukuqonda imitapo yolwazi ehlobene yePython
Lapho sisebenza ngama-tensor namamodeli okufunda omshini, sivamise ukusebenzisa imitapo yolwazi eminingi yePython.
I-TensorFlow kanye ne -PyTorch ziyilabhulali ezimbili ezidumile kakhulu zemisebenzi eyahlukahlukene yokufunda komshini kanye nokufunda okujulile. Zisekela imisebenzi eminingi ye-tensor futhi ziza nezici eziningi ezakhelwe ngaphakathi zokwakha nokuqeqesha amanethiwekhi ezinzwa.
import tensorflow as tf # Creating a tensor using TensorFlow tensor_tf = tf.constant([[1, 2], [3, 4]]) print(tensor_tf) # Creating a tensor using PyTorch import torch tensor_torch = torch.tensor([[1, 2], [3, 4]]) print(tensor_torch)
I-NumPy ingenye ilabhulali enamandla yokubala izinombolo. Nakuba ingaklanyelwanga ngokusobala ukufunda okujulile njenge-TensorFlow ne-PyTorch, isetshenziswa kakhulu ekuphatheni idatha yezinombolo ngenxa yokusebenza kwayo kahle kanye nokulula kokusetshenziswa.
import numpy as np # Creating a tensor (in this case, a multi-dimensional array) using NumPy tensor_np = np.array([[1, 2], [3, 4]]) print(tensor_np)
Ukwembula Ukusebenza Kwekhodi
Ngemva kokulungiselela okokufaka kwe-tensor, isinyathelo esilandelayo kumodeli yethu yokubikezela ukuklama imodeli ngokwayo. Ukwenza kube lula, ake sithi sifuna ukusebenzisa imodeli yomugqa wokuhlehla. Ku-TensorFlow, singakwenza lokhu kalula nje.
# define a simple linear regression model
model = tf.keras.Sequential([
tf.keras.layers.Dense(units=1, input_shape=[1])
])
# compile the model
model.compile(optimizer=tf.keras.optimizers.RMSprop(0.01), loss='mean_squared_error', metrics=['mean_absolute_error'])
#fit the model to the data
history = model.fit(input_tensor, epochs=500, verbose=0)
Kule khodi engenhla, imodeli elula yomugqa yokuhlehla yenziwa kusetshenziswa i-TensorFlow's Keras API bese iyahlanganiswa. I-'RMSProp' i-algorithm yokuthuthukisa esiyisebenzisela ukwenza imodeli ibe ngcono ngesikhathi sokuqeqeshwa, futhi 'i-mean_squared_error' umsebenzi wokulahlekelwa, indlela yezibalo yokulinganisa ukuthi izibikezelo zemodeli azilungile kangakanani. Ngemuva kokusetha imodeli, siyiqeqesha sisebenzisa umsebenzi 'wokufana'. Inqubo yokuqeqesha ibandakanya ukuphakela i-tensor yokokufaka kumodeli nokuyivumela ukuthi yenze izibikezelo. Izisindo zemodeli ziyalungiswa ukuze izibikezelo zibe nembe ngangokunokwenzeka.
Ngokuqonda okuyisisekelo kwe-Tensors, imitapo yolwazi ye-Python, kanye nokusebenza kwekhodi, abafundi bangathatha lolu lwazi ukuze bazilolonge, ngaleyo ndlela basize ukuvala igebe phakathi kokuqonda kwethiyori nokungokoqobo esikhaleni sohlelo lwePython nokufunda komshini.
