Conditional GANs Quiz

Answers and explanations appear as you take the quiz.

1. What is the main difference between a conditional GAN and a standard GAN?
   • It feeds extra information like a class label into both networks to control output
   • It uses two generators
   • It removes the discriminator
   • It only works on text, not images
2. In a cGAN trained on labeled digits, what can you do that a plain GAN cannot?
   • Run without any noise input
   • Request a specific digit such as a '7'
   • Skip the discriminator entirely
   • Generate only black-and-white noise
3. What must the cGAN discriminator check, beyond whether an image looks real?
   • The file format of the image
   • The resolution in megapixels
   • Whether the image matches its given label
   • The training learning rate
4. Which is a common way to supply the conditioning signal to the generator?
   • Deleting the noise vector
   • Training without any labels
   • Using a separate physical sensor
   • Concatenating the label to the noise input
5. Conditional GANs laid the groundwork for which later capability?
   • Text-to-image and image-to-image generation
   • Faster CPUs
   • Spam email filtering
   • Database indexing
6. In modern diffusion models, the cGAN idea of conditioning often appears as what mechanism?
   • Random weight clipping
   • Cross-attention on a text prompt
   • Manual pixel painting only
   • Disabling the noise schedule
7. Why must a conditional discriminator see both the generated image and its conditioning label?
   • It uses the label to choose the learning-rate schedule
   • It converts the label into a random noise seed only
   • It needs to judge whether the image is realistic and whether it matches the requested condition
   • It prevents the generator from receiving noise
8. Besides concatenating a class label to the noise vector, which conditioning method is described in the guide?
   • A spectral-radius constraint
   • An inverse Hessian update
   • A sequence-level BLEU reward
   • A projection layer that scores compatibility between label and image features