Question $1.$ Compared to shallow nets, deep neural networks are able to:

a$.$ Reach much better performance with the same amount of training data

b$.$ Reach much worse performance with the same amount of training data

c$.$ Have a much lower parameter efficiency for complex problems

d$.$ Model complex functions using exponentially more parameters

Explain your answer $.$ Question $2$:

Which activation function is used in a neural network to transform the network's raw output scores $($logits$)$ into a probability distribution over multiple classes?

a$.$ Sigmoid

b$.$ ReLU

c$.$ Hyperbolic Tangent $($tanh$)$

d$.$ Softmax

Explain your answer. Question $3$:

Which statement accurately describes the backpropagation algorithm in a Deep Neural Network $($DNN$)?$

a$.$ Backpropagation requires as many passes through the network as the number of neurons in the input layer

b$.$ Backpropagation uses a forward pass to make a prediction and measure error, followed by a reverse pass to compute gradients and update weights

c$.$ Backpropagation performs Gradient Descent in a single forward pass through the network to minimize error

d$.$ Backpropagation computes gradients for only a subset of model parameters to speed up training

Explain your answer. Question $4$:

What type of machine learning is self$-$supervised learning considered to be$?$

a$.$ Supervised Learning

b$.$ Unsupervised Learning

c$.$ Reinforcement Learning

d$.$ Instance$-$based Learning

Explain your answer. Question $5$:

Which of the following is not a benefit of pooling layers in CNN$?$

a$.$ Reducing the computational load

b$.$ Reducing the memory usage

c$.$ Reducing the number of parameters

d$.$ Reducing the risk of underfitting

Explain your answer