IBM AI Engineering Neural Networks with Keras Quiz

1. What is the difference between Sequential and Functional API in Keras?

Sequential is faster; Functional is slower

Sequential builds linear models; Functional supports complex architectures

Functional is only for CNNs

Sequential requires more memory

2. Which weight initialization method is commonly used in deep neural networks?

Xavier initialization

He initialization

Random normal distribution

Both A and B are widely used

3. In Keras, what is the purpose of the validation_split parameter in fit()?

To reserve a portion of training data for validation

To split test data into batches

To divide the dataset geographically

To create separate models for each class

4. What does the softmax activation function produce in a neural network?

A single probability value

Binary outputs (0 or 1)

A probability distribution over multiple classes

Unbounded real values

5. Which Keras callback can be used to stop training when validation loss stops improving?

ModelCheckpoint

EarlyStopping

ReduceLROnPlateau

TensorBoard

6. What is the role of the loss function during model training?

To evaluate model performance on test data

To measure the difference between predictions and actual values

To initialize network weights

To determine the number of layers

7. In Keras, what metric is commonly used for multi-class classification problems?

Binary accuracy

Categorical accuracy

Mean squared error

Precision

8. What does batch normalization do in a neural network?

Splits data into batches

Normalizes layer inputs to stabilize training

Reduces the number of parameters

Increases model interpretability

9. Which Keras layer type performs max pooling for dimensionality reduction?

MaxPooling2D

AveragePooling2D

GlobalMaxPooling2D

All of the above

10. In the context of neural networks, what does 'epoch' mean?

One forward and backward pass through all training data

A single training example

A hyperparameter adjustment

A validation checkpoint

11. What is the primary purpose of an activation function in a neural network?

To initialize network weights

To introduce non-linearity into the model

To reduce memory consumption

To normalize input data

12. Which activation function outputs values between 0 and 1?

ReLU

Sigmoid

Tanh

Linear

13. In Keras, what is the purpose of the fit() method?

To test the model on unseen data

To compile the model

To train the model on training data

To reshape input tensors

14. What does a Conv2D layer in Keras perform?

2D convolution for image processing

Vector multiplication

Sequence padding

Data normalization

15. Which optimizer is commonly used for training deep neural networks in Keras?

Gradient Descent

Adam

Stochastic Gradient Descent

Both B and C are correct

16. In neural networks, what does backpropagation accomplish?

Initializes network weights randomly

Computes gradients and updates weights during training

Normalizes input features

Splits data into train and test sets

17. What is the purpose of the Dropout layer in Keras?

To accelerate training

To prevent overfitting by randomly deactivating neurons

To normalize activations

To increase model accuracy

18. Which activation function is typically used in the output layer for binary classification?

ReLU

Sigmoid

Tanh

Softmax

19. In Keras, what does the compile() method define?

The input shape of data

The optimizer, loss function, and metrics

The number of training epochs

The batch size for training

20. Which Keras layer type is most commonly used for processing sequential data?

Dense

LSTM

Conv2D

Dropout