10)Build a Convolution Neural Network for MNIST Hand written Digit Classification. Code: importkeras fromkeras.models import Sequential fromkeras.layers import Dense,Conv2D,MaxPooling2D,Flatten fromkeras.datasets import mnist fromkeras.utils import to_categorical frommatplotlib import pyplot as plt (xt, yt), (xte, yte) = mnist.load_data() '''for i in range(20): plt.subplot(5,4,i+1) plt.imshow(xt[i]) plt.title("Digits:{}".format(yt[i])) plt.subplots_adjust(hspace=0.5) plt.axis('off')''' # Preprocess the data xt = xt.reshape((xt.shape[0], 28, 28)).astype('float32') / 255 xte = xte.reshape((xte.shape[0], 28, 28)).astype('float32') / 255 yt = to_categorical(yt) yte = to_categorical(yte) # Define the CNN model n=Sequential() n.add(Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1))) n.add(MaxPooling2D((2, 2))) n.add(Conv2D(64, (3, 3), activation='relu')) n.add(MaxPooling2D((2, 2))) n.add(Conv2D(64, (3, 3), activation='relu')) n.add(Flatten()) n.add(Dense(64, activation='relu')) n.add(Dense(10, activation='softmax')) # Compile the model n.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['accuracy']) # Train the modelhistory = importnumpy as np n.fit(xt, yt,epochs=2) '''plt.imshow(xte[0]) plt.title("Actual value:{}".format(yte[0]))''' p=n.predict(xte) plt.axis('off') ifnp.argmax(yte[0])==np.argmax(p[0]): print("correct p") else: print("incorrect cp") # Evaluate the model test_loss, test_acc = n.evaluate(xte, yte) print('Test accuracy:', test_acc) Output: Test accuracy:97.7