TensorFlow Neural Network Construction Quiz

1. In TensorFlow's Functional API, how are models defined differently from Sequential models?

By stacking layers sequentially

By using Input objects and connecting layers through function calls

By inheriting from tf.keras.Model

By defining layers in a dictionary

2. What is the primary advantage of using callbacks in TensorFlow training?

Increases computational speed

Monitors and controls training behavior dynamically

Automatically selects optimal hyperparameters

Reduces memory consumption

3. Which regularization technique adds a penalty term to the loss function?

Dropout

Data augmentation

L1 or L2 regularization

Batch normalization

4. In TensorFlow, what does the epochs parameter control during training?

The number of samples per batch

The total number of passes through the entire dataset

The learning rate adjustment schedule

The fraction of data used for validation

5. What is the effect of using ReLU activation in hidden layers?

Introduces non-linearity and avoids vanishing gradients

Squashes outputs to a fixed range

Reduces model complexity

Normalizes feature distributions

6. Which approach allows you to define custom layers and training logic in TensorFlow?

Using Sequential API only

Subclassing tf.keras.Model

Using only pre-built layers

Defining layers in configuration files

7. In TensorFlow, what does the summary() method provide?

Training history and loss curves

A detailed overview of model architecture and parameters

Prediction results on test data

Feature importance rankings

8. What is the purpose of the validation_split parameter in the fit() method?

Splits data into training and test sets

Reserves a portion of training data to evaluate model performance

Divides the dataset into batches

Separates features from labels

9. Which metric is most appropriate for evaluating a binary classification model?

Mean Squared Error

Accuracy

Mean Absolute Error

Cosine Similarity

10. What does BatchNormalization do to layer inputs?

Removes duplicate samples

Normalizes inputs to zero mean and unit variance per batch

Converts categorical data to numerical

Reduces feature dimensionality

11. Which TensorFlow API is most commonly used for building sequential neural networks?

Tf.keras.Sequential

Tf.estimator.Estimator

Tf.raw_ops

Tf.compat.v1.layers

12. Which loss function is appropriate for multi-class classification problems?

Binary_crossentropy

Sparse_categorical_crossentropy

Mean_squared_error

Hinge

13. What is the typical role of Dropout in neural network construction?

Increases model capacity

Prevents overfitting by randomly deactivating neurons

Accelerates training speed

Normalizes layer outputs

14. In TensorFlow, which optimizer is known for adaptive learning rates and is commonly used in modern networks?

SGD

Momentum

Adam

Adagrad

15. What does the Flatten layer do in a neural network?

Removes outliers from the data

Converts multidimensional input into a 1D vector

Reduces overfitting through regularization

Normalizes feature values

16. Which layer type is used to reduce spatial dimensions while preserving important features?

MaxPooling2D

Flatten

BatchNormalization

Dropout

17. What is the output shape of a Conv2D layer with 32 filters, 3×3 kernel, and 'valid' padding on a 28×28×1 input?

26×26×32

28×28×32

30×30×32

25×25×32

18. In TensorFlow, what does the compile() method configure for a model?

The number of training epochs

Optimizer, loss function, and metrics

The batch size for training

The validation split ratio

19. Which activation function is best suited for the output layer in binary classification?

ReLU

Sigmoid

Tanh

Linear

20. What is the primary purpose of the Dense layer in neural networks?

Applies dropout to reduce overfitting

Connects every input neuron to every output neuron with weights

Normalizes input data to zero mean and unit variance

Performs 2D convolution operations

TensorFlow Developer Neural Network Construction Quiz