AI AI Engineer: Neural Networks and Deep Learning

Question 1

What is the vanishing gradient problem in deep neural networks?

Accepted Answer

Gradients become extremely small, making early layers learn very slowly

Answer

Gradients become too large during backpropagation

Answer

The loss function fails to converge

Answer

Weights are initialized to zero

Question 2

Which activation function is most commonly used in hidden layers of modern deep neural networks to mitigate the vanishing gradient problem?

Accepted Answer

ReLU (Rectified Linear Unit)

Answer

Sigmoid

Answer

Tanh

Answer

Softmax

Question 3

What is the purpose of batch normalization in a neural network?

Accepted Answer

To normalize activations within a layer to stabilize and accelerate training

Answer

To split data into mini-batches

Answer

To regularize weights using L2 penalty

Answer

To reduce the number of parameters

Question 4

In a convolutional neural network (CNN), what does 'stride' control?

Accepted Answer

How many pixels the filter moves at each step

Answer

The depth of the network

Answer

The number of filters applied

Answer

The size of the pooling window

Question 5

What is dropout regularization in neural networks?

Accepted Answer

Randomly setting a fraction of neurons to zero during training to prevent overfitting

Answer

Removing layers with poor performance

Answer

Reducing learning rate during training

Answer

Pruning weights below a threshold

Question 6

Which optimizer adapts the learning rate for each parameter individually using estimates of first and second moments of gradients?

Accepted Answer

Adam

Answer

SGD (Stochastic Gradient Descent)

Answer

Momentum

Answer

Adagrad