Virtual Labs

Generative Adversarial Networks (GANs)

1. What are the two main components of a GAN?

a: Generator and encoder Explanation

Explanation

b: Discriminator and encoder Explanation

Explanation

c: Generator and discriminator Explanation

Explanation

d: Encoder and discriminator Explanation

Explanation

2. How does the discriminator improve during GAN training?

a: By learning to classify real and generated samples correctly Explanation

Explanation

b: By directly updating the generator weights Explanation

Explanation

c: By reducing the dimensionality of the latent space Explanation

Explanation

d: By generating synthetic samples Explanation

Explanation

3. Which loss function is commonly used in standard GANs?

a: Hinge loss Explanation

Explanation

b: Mean squared error loss Explanation

Explanation

c: Binary cross-entropy (logistic loss) Explanation

Explanation

d: Kullback-Leibler divergence Explanation

Explanation

4. What is the purpose of the latent space in a GAN?

a: To represent the high-dimensional space of real samples Explanation

Explanation

b: To control the diversity and characteristics of generated samples Explanation

Explanation

c: To adjust the learning rate during training Explanation

Explanation

d: None of the above Explanation

Explanation

5. How is GAN training formulated in the original GAN framework proposed by Ian Goodfellow?

a: The generator and discriminator are trained independently Explanation

Explanation

b: The generator and discriminator are trained simultaneously in a minimax game Explanation

Explanation

c: The generator is fully trained before the discriminator Explanation

Explanation

d: The discriminator is permanently fixed during training Explanation

Explanation

6. Which technique is commonly used to improve GAN training stability?

a: Learning rate scheduling Explanation

Explanation

b: Increasing image resolution at every iteration Explanation

Explanation

c: Removing the discriminator network Explanation

Explanation

d: Disabling backpropagation Explanation

Explanation

7. What does it indicate when the discriminator can no longer reliably distinguish real from generated samples?

a: The GAN has reached a balanced adversarial state Explanation

Explanation

b: The discriminator has stopped updating permanently Explanation

Explanation

c: The generator has failed completely Explanation

Explanation

d: The discriminator becomes the generator Explanation

Explanation

8. What is the purpose of the adversarial loss in GANs?

a: To encourage the generator to produce samples that deceive the discriminator Explanation

Explanation

b: To control the learning rate during training Explanation

Explanation

c: To adjust the weights and biases of the generator Explanation

Explanation

d: None of the above Explanation

Explanation

9. What is mode collapse in GAN training?

a: The discriminator stops learning completely Explanation

Explanation

b: The generator produces limited varieties of outputs repeatedly Explanation

Explanation

c: The latent space becomes larger during training Explanation

Explanation

d: The discriminator generates synthetic images Explanation

Explanation

10. Which technique can help reduce mode collapse in GANs?

a: Gradient penalty Explanation

Explanation

b: Batch normalization Explanation

Explanation

c: Dropout Explanation

Explanation

d: All of the above Explanation

Explanation