Convolutional Neural Networks in TensorFlow Coursera Exercise Quiz Answers

 

Week 1

Question 1: What does flow_from_directory give you on the ImageGenerator?

• The ability to easily load images for training
• The ability to pick the size of training images
• The ability to automatically label images based on their directory name
• All of the above

Question 2: If my Image is sized 150×150, and I pass a 3×3 Convolution over it, what size is the resulting image?

• 148×148
• 150×150
• 153×153
• 450×450

Question 3: If my data is sized 150×150, and I use Pooling of size 2×2, what size will the resulting image be?

• 300×300
• 148×148
• 149×149
• 75×75

Question 4: If I want to view the history of my training, how can I access it?

• Create a variable ‘history’ and assign it to the return of model.fit or model.fit_generator
• Pass the parameter ‘history=true’ to the model.fit
• Use a model.fit_generator
• Download the model and inspect it

Question 5: What’s the name of the API that allows you to inspect the impact of convolutions on the images?

• The model.pools API
• The model.layers API
• The model.images API
• The model.convolutions API

Question 6: When exploring the graphs, the loss levelled out at about .75 after 2 epochs, but the accuracy climbed close to 1.0 after 15 epochs. What’s the significance of this?

• There was no point training after 2 epochs, as we overfit to the validation data
• There was no point training after 2 epochs, as we overfit to the training data
• A bigger training set would give us better validation accuracy
• A bigger validation set would give us better training accuracy

Question 7: Why is the validation accuracy a better indicator of model performance than training accuracy?

• It isn’t, they’re equally valuable
• There’s no relationship between them
• The validation accuracy is based on images that the model hasn’t been trained with, and thus a better indicator of how the model will perform with new images.
• The validation dataset is smaller, and thus less accurate at measuring accuracy, so its performance isn’t as important

Question 8: Why is overfitting more likely to occur on smaller datasets?

• Because in a smaller dataset, your validation data is more likely to look like your training data
• Because there isn’t enough data to activate all the convolutions or neurons
• Because with less data, the training will take place more quickly, and some features may be missed
• Because there’s less likelihood of all possible features being encountered in the training process.

Week 2

Question 1: How do you use Image Augmentation in TensorFLow

• Using parameters to the ImageDataGenerator
• With the keras.augment API
• You have to write a plugin to extend tf.layers
• With the tf.augment API

Question 2: If my training data only has people facing left, but I want to classify people facing right, how would I avoid overfitting?

• Use the ‘horizontal_flip’ parameter
• Use the ‘flip’ parameter and set ‘horizontal’
• Use the ‘flip’ parameter
• Use the ‘flip_vertical’ parameter around the Y axis

Question 3: When training with augmentation, you noticed that the training is a little slower. Why?

• Because the augmented data is bigger
• Because the image processing takes cycles
• Because there is more data to train on
• Because the training is making more mistakes

Question 4: What does the fill_mode parameter do?

• There is no fill_mode parameter
• It creates random noise in the image
• It attempts to recreate lost information after a transformation like a shear
• It masks the background of an image

Question 5: When using Image Augmentation with the ImageDataGenerator, what happens to your raw image data on-disk.

• It gets overwritten, so be sure to make a backup
• A copy is made and the augmentation is done on the copy
• Nothing, all augmentation is done in-memory
• It gets deleted

Question 6: How does Image Augmentation help solve overfitting?

• It slows down the training process
• It manipulates the training set to generate more scenarios for features in the images
• It manipulates the validation set to generate more scenarios for features in the images
• It automatically fits features to images by finding them through image processing techniques

Question 7: When using Image Augmentation my training gets…

• Slower
• Faster
• Stays the Same
• Much Faster

Question 8: Using Image Augmentation effectively simulates having a larger data set for training.

• False
• True

Week 3

Question 1: If I put a dropout parameter of 0.2, how many nodes will I lose?

• 20% of them
• 2% of them
• 20% of the untrained ones
• 2% of the untrained ones

Question 2: Why is transfer learning useful?

• Because I can use all of the data from the original training set
• Because I can use all of the data from the original validation set
• Because I can use the features that were learned from large datasets that I may not have access to
• Because I can use the validation metadata from large datasets that I may not have access to

Question 3: How did you lock or freeze a layer from retraining?

• tf.freeze(layer)
• tf.layer.frozen = true
• tf.layer.locked = true
• layer.trainable = false

Question 4: How do you change the number of classes the model can classify when using transfer learning? (i.e. the original model handled 1000 classes, but yours handles just 2)

• Ignore all the classes above yours (i.e. Numbers 2 onwards if I’m just classing 2)
• Use all classes but set their weights to 0
• When you add your DNN at the bottom of the network, you specify your output layer with the number of classes you want
• Use dropouts to eliminate the unwanted classes

Question 5: Can you use Image Augmentation with Transfer Learning Models?

• No, because you are using pre-set features
• Yes, because you are adding new layers at the bottom of the network, and you can use image augmentation when training these

Question 6: Why do dropouts help avoid overfitting?

• Because neighbor neurons can have similar weights, and thus can skew the final training
• Having less neurons speeds up training

Question 7: What would the symptom of a Dropout rate being set too high?

• The network would lose specialization to the effect that it would be inefficient or ineffective at learning, driving accuracy down
• Training time would increase due to the extra calculations being required for higher dropout

Question 8: Which is the correct line of code for adding Dropout of 20% of neurons using TensorFlow

• tf.keras.layers.Dropout(20)
• tf.keras.layers.DropoutNeurons(20),
• tf.keras.layers.Dropout(0.2),
• tf.keras.layers.DropoutNeurons(0.2),

Week 4

Question 1: The diagram for traditional programming had Rules and Data In, but what came out?

• Answers
• Binary
• Machine Learning
• Bugs

Question 2: Why does the DNN for Fashion MNIST have 10 output neurons?

• To make it train 10x faster
• To make it classify 10x faster
• Purely Arbitrary
• The dataset has 10 classes

Question 3: What is a Convolution?

• A technique to make images smaller
• A technique to make images larger
• A technique to extract features from an image
• A technique to remove unwanted images

Question 4: Applying Convolutions on top of a DNN will have what impact on training?

• It will be slower
• It will be faster
• There will be no impact
• It depends on many factors. It might make your training faster or slower, and a poorly designed Convolutional layer may even be less efficient than a plain DNN!

Question 5: What method on an ImageGenerator is used to normalize the image?

• normalize
• flatten
• rezize()
• rescale

Question 6: When using Image Augmentation with the ImageDataGenerator, what happens to your raw image data on-disk.

• A copy will be made, and the copies are augmented
• A copy will be made, and the originals will be augmented
• Nothing
• The images will be edited on disk, so be sure to have a backup

Question 7: Can you use Image augmentation with Transfer Learning?

• No – because the layers are frozen so they can’t be augmented
• Yes. It’s pre-trained layers that are frozen. So you can augment your images as you train the bottom layers of the DNN with them

Question 8: When training for multiple classes what is the Class Mode for Image Augmentation?

• class_mode=’multiple’
• class_mode=’non_binary’
• class_mode=’categorical’
• class_mode=’all’