$2\_$final.pdf

of Code

Leetcode

ChatGPT

jupyter

Perplexity

$72\%$

$($f$)(3$ pts$)$ We want to train a GAN for generating fake images. Ideally, at the end of training, the discriminator's classification accuracy should be $q,$

A$.$ almost $0\%.$

B$.$ around $50\%.$

C$.$ somewhere between $50\%$ and $100\%.$

D$.$ almost $100\%.$

$($g$)(3$ pts$)$ For a video game, the action space is $\{"$left$",$ "right", $"$up$"\}.$ The policy function predicts:

Which of the following statements is correct?

A$.$ "left" is the agent's action $a_t.$

B$.$ "right" is the agent's action $a_t.$

C$."$up$"$ is the agent's action $a_t.$

D$.$ Any of "left", "right", and $"$up$"$ can be the agent's action $a_t.$

$($h$)(3$ pts$)$ Suppose we seek to predict housing price $($a positive number$)$ using a multi$-$layer perceptron. Which of the followings is the best choice for the activation function of the output layer?

A$.$ No activation function $($identity function$).$

B$.$ ReLU.

C$.$ Sigmoid function.

D$.$ Softmax function.

$($i$)(3$ pts$)$ Let $h_t$ be the hidden state of a SimpleRNN model. What is the range of the entries of $h_t?($Choose the most precise one from followings.$)$

A$.In(-,+).$

B$.$ In $(-1,1).$

C$.$ In $(0,1).$

D$.ln(0,+).$

$($i$)(3$ pts$)$ Which of the following functions can be used to control the agent?

A$.$ Action$-$value function $Q_{-}(s,a).$

B$.$ Optimal action$-$value function $Q^{**}(a,a).$

C$.$ Statevalue function $V_5(o).$

D$.$ Policy function $(s|).$

Page $4$