MNIST
Table of Contents
1. MNIST
import tensorflow as tf import numpy as np from tensorflow.keras.datasets import mnist from tensorflow.keras import layers,losses,metrics,optimizers, models # load data
1.1. simple
(x_train,y_train), (x_test, y_test) = mnist.load_data() tf.keras.backend.clear_session() model = models.Sequential() model.add(layers.Flatten(input_shape = (28, 28))) model.add(layers.Dense(100, activation = "relu")) model.add(layers.Dense(10, activation = "softmax")) model.compile(optimizer="adam",loss="sparse_categorical_crossentropy", metrics=["accuracy"]) model.fit(x_train, y_train, batch_size = 200, epochs = 10, verbose = 0) model.evaluate(x_test, y_test, verbose=2)
313/313 - 0s - loss: 0.3964 - accuracy: 0.9426
1.2. one-hot
(x_train,y_train), (x_test, y_test) = mnist.load_data() tf.keras.backend.clear_session() x_train = tf.reshape(x_train, (-1, 28 * 28)) y_train = tf.one_hot(y_train, depth = 10) x_test = tf.reshape(x_test, (-1, 28 * 28)) y_test = tf.one_hot(y_test, depth = 10) model = models.Sequential() model.add(layers.Dense(100, input_shape = (28 * 28, ), activation = "relu")) model.add(layers.Dense(10, activation = "softmax")) model.compile(optimizer="adam",loss="categorical_crossentropy", metrics=["accuracy"]) model.fit(x_train, y_train, batch_size = 200, epochs = 10, verbose = 0) model.evaluate(x_test, y_test, verbose=2)
313/313 - 0s - loss: 0.3311 - accuracy: 0.9362
1.3. manually fit
(x_train,y_train), (x_test, y_test) = mnist.load_data()
tf.keras.backend.clear_session()
x_train = tf.reshape(x_train, (-1, 28 * 28))
y_train = tf.one_hot(y_train, depth = 10)
x_test = tf.reshape(x_test, (-1, 28 * 28))
y_test = tf.one_hot(y_test, depth = 10)
model = models.Sequential()
model.add(layers.Dense(100, input_shape = (28 * 28, ), activation = "relu"))
model.add(layers.Dense(10, activation = "softmax"))
loss_func = tf.keras.losses.CategoricalCrossentropy()
evaluator = tf.keras.metrics.CategoricalAccuracy(name='test_accuracy')
optimizer = tf.keras.optimizers.Adam()
train_dataset = tf.data.Dataset.from_tensor_slices(
(x_train, y_train)).batch(200)
for epoch in range(20):
for x, y in train_dataset:
with tf.GradientTape() as tape:
y_hat= model(x)
loss = loss_func(y, y_hat)
gradients = tape.gradient(loss, model.trainable_variables)
optimizer.apply_gradients(zip(gradients, model.trainable_variables))
evaluator(y_test, model(x_test))
print(evaluator.result())
tf.Tensor(0.955, shape=(), dtype=float32)
