Scribd
Upload
6 views8 pages

DL Week-7-1

Week 7

Uploaded by

shruthivatham33
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views8 pages

DL Week-7-1

Week 7

Uploaded by

shruthivatham33
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

Week-7: consider the network and dataset from week 6 visualize the

hidden layers features. Compute the confusion matrix.


Code:
import numpy as np
import tensorflow as tf
from tensorflow.keras import layers, models
from sklearn.model_selection import KFold
from tensorflow.keras.datasets import mnist
from sklearn.metrics import accuracy_score, confusion_matrix
import matplotlib.pyplot as plt

# Load the MNIST dataset


(x_train, y_train), (x_test, y_test) = mnist.load_data()

# Normalize pixel values to be between 0 and 1


x_train, x_test = x_train / 255.0, x_test / 255.0

# Expand dimensions to add a channel dimension


x_train = np.expand_dims(x_train, axis=-1)
x_test = np.expand_dims(x_test, axis=-1)

# Define the CNN model with 4 layers


def create_model():
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28,
28, 1)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10, activation='softmax'))
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
return model

# Print model summary


model = create_model()
model.summary()

# Perform K-fold cross-validation


k_folds = 10
kf = KFold(n_splits=k_folds, shuffle=True, random_state=42)

accuracies = []
conf_matrices = []

for train_index, val_index in kf.split(x_train):


x_train_fold, x_val_fold = x_train[train_index], x_train[val_index]
y_train_fold, y_val_fold = y_train[train_index], y_train[val_index]
model = create_model()

# Train the model


model.fit(x_train_fold, y_train_fold, epochs=5, batch_size=64, verbose=0)

# Evaluate on the validation set


val_predictions = np.argmax(model.predict(x_val_fold), axis=-1)
accuracy = accuracy_score(y_val_fold, val_predictions)
accuracies.append(accuracy)

# Compute confusion matrix


conf_matrix = confusion_matrix(y_val_fold, val_predictions)
conf_matrices.append(conf_matrix)

# Print the accuracy over the K fold


print('Accuracy is:', accuracy)

# Print the confusion matrix


for fold, conf_matrix in enumerate(conf_matrices):
print(f'Confusion Matrix for Fold {fold + 1}:')
print(conf_matrix)
# Print the average accuracy over the K folds
print(f'Average Accuracy: {np.mean(accuracies)}')

# Visualize hidden layer features


layer_outputs = [layer.output for layer in model.layers[:4]] # Extracting
the outputs of the top 4 layers
activation_model = tf.keras.models.Model(inputs=model.input,
outputs=layer_outputs) # Creates a model that will return these outputs,
given the model input

sample_img = x_train[0].reshape(1, 28, 28, 1) # Reshape an input


sample
activations = activation_model.predict(sample_img) # Predictions of
the activations

# Plot the activations of each layer


for layer_activation, layer_name in zip(activations, ['conv1', 'maxpool1',
'conv2', 'flatten']):
plt.figure()
plt.matshow(layer_activation[0, :, :, 0], cmap='viridis')
plt.title(layer_name)
plt.colorbar()

plt.show()
Output:
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 26, 26, 32) 320

max_pooling2d (MaxPooling2 (None, 13, 13, 32) 0


D)

conv2d_1 (Conv2D) (None, 11, 11, 64) 18496

flatten (Flatten) (None, 7744) 0

dense (Dense) (None, 64) 495680

dense_1 (Dense) (None, 10) 650

=================================================================
Total params: 515146 (1.97 MB)
Trainable params: 515146 (1.97 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

188/188 [==============================] - 4s 17ms/step


Accuracy is: 0.9588333333333333
188/188 [==============================] - 4s 16ms/step
Accuracy is: 0.9598333333333333
188/188 [==============================] - 4s 16ms/step
Accuracy is: 0.9616666666666667
188/188 [==============================] - 3s 14ms/step
Accuracy is: 0.9663333333333334
188/188 [==============================] - 3s 13ms/step
Accuracy is: 0.9595
188/188 [==============================] - 4s 16ms/step
Accuracy is: 0.972
188/188 [==============================] - 3s 15ms/step
Accuracy is: 0.96
188/188 [==============================] - 3s 13ms/step
Accuracy is: 0.9571666666666667
188/188 [==============================] - 3s 13ms/step
Accuracy is: 0.9611666666666666
188/188 [==============================] - 5s 19ms/step
Accuracy is: 0.9651666666666666
Confusion Matrix for Fold 1:
[[604 0 3 2 0 2 3 1 8 1]
[ 0 645 2 0 0 0 0 0 6 1]
[ 0 2 559 2 2 0 0 2 4 1]
[ 2 1 7 557 0 10 0 1 11 0]
[ 1 3 3 2 562 0 1 1 4 3]
[ 2 2 2 9 3 520 2 0 11 0]
[ 2 0 1 0 2 4 568 0 3 0]
[ 1 6 12 3 3 3 0 589 10 6]
[ 1 4 1 3 1 1 4 0 570 0]
[ 1 2 1 9 19 3 0 3 15 579]]
Confusion Matrix for Fold 2:
[[545 0 1 0 2 1 0 0 0 2]
[ 0 659 4 2 1 0 0 0 1 1]
[ 0 5 584 2 1 0 0 6 3 1]
[ 1 1 5 602 0 9 0 3 4 5]
[ 0 0 0 0 583 0 1 0 0 12]
[ 5 1 1 12 1 524 3 0 1 5]
[ 0 3 2 0 5 8 576 0 3 0]
[ 1 2 4 4 6 1 0 634 0 14]
[ 2 6 4 23 4 8 2 3 512 11]
[ 2 2 0 1 8 3 0 6 0 540]]
-
-
-
-
-
-
-
Confusion Matrix for Fold 10:
[[578 0 1 0 1 2 5 1 1 0]
[ 0 663 3 0 0 1 1 2 4 0]
[ 1 1 573 0 5 0 0 4 4 1]
[ 0 3 9 539 0 15 1 4 7 0]
[ 1 0 1 0 566 0 4 2 0 10]
[ 0 1 0 3 2 518 3 0 1 2]
[ 7 4 1 1 0 7 586 0 3 0]
[ 0 1 3 0 1 1 1 609 1 6]
[ 1 3 4 1 1 7 4 1 599 4]
[ 0 1 2 4 5 9 0 13 5 560]]

Average Accuracy: 0.9621666666666666


1/1 [==============================] - 1s 772ms/step

You might also like