Machine Learning Foundations: Exercise 3 Improve accuracy of MNIST using Convolutions.

Exercise 3: Improve accuracy of MNIST using Convolutions:code lab link

Improve MNIST to 99.8% accuracy or more using only a single convolutional layer and a single MaxPooling 2D.

The filter amount will affect the accuracy and training time.

Code: 

	import tensorflow as tf
	# Callback function to check model accuracy
	class RayCallback(tf.keras.callbacks.Callback):
	def on_epoch_end(self, epoch, logs={}):
	if(logs.get('accuracy')>0.998):
	print("\nReached 99.8% accuracy so cancelling training!")
	self.model.stop_training = True
	# Load the MNIST handwrite digit data set
	mnist = tf.keras.datasets.mnist
	(training_images, training_labels), (test_images, test_labels) = mnist.load_data()
	# Reshap and normalize training data and callback function
	callbacks = RayCallback()
	training_images=training_images.reshape(60000, 28, 28, 1)
	training_images = training_images/255.0
	# Create an 5 layer model:
	# Convolution 16 filter with 3 X 3 size to each Image -> Polling each Image to 1/4
	# Flatten -> 128 input -> 10 output
	model = tf.keras.models.Sequential([
	tf.keras.layers.Conv2D(16, (3,3), activation='relu', input_shape=(28, 28, 1)),
	tf.keras.layers.MaxPooling2D(2, 2),
	tf.keras.layers.Flatten(),
	tf.keras.layers.Dense(128, activation='relu'),
	tf.keras.layers.Dense(10, activation='softmax')
	])
	# Setting optimizer and loss function
	model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
	# Training model untill accuracy > 99.8%
	model.fit(training_images, training_labels, epochs=10, callbacks=[callbacks])

view raw gistfile1.txt hosted with ❤ by GitHub