wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

Generative AI MCQ Quiz

Total questions: 75

Worksheet time: 48mins

Name
Class
Date
1.

What is the primary objective of a GAN's generator?

a)

To classify images

b)

To detect anomalies

c)

To generate realistic data that can fool the discriminator

d)

To calculate loss functions

2.

Which of the following is NOT a role of the discriminator in a GAN?

a)

Distinguish real data from fake data

b)

Provide gradient feedback to the generator

c)

Generate new data samples

d)

Improve over time to detect better fakes

3.

What kind of loss is used in the original GAN paper for the generator?

a)

Mean Squared Error

b)

Binary Cross-Entropy

c)

Hinge Loss

d)

Wasserstein Loss

4.

What is 'mode collapse' in GANs?

a)

The generator learns the entire data distribution

b)

The discriminator becomes too weak

c)

The generator produces limited types of outputs

d)

The training fails due to overfitting

5.

Which of the following helps in improving GAN training stability?

a)

Decreasing the number of epochs

b)

Using ReLU in the discriminator

c)

One-sided label smoothing

d)

Removing the discriminator

6.

What architecture does DCGAN use in place of fully connected layers?

a)

RNN layers

b)

LSTM layers

c)

Convolutional layers

d)

Transformer blocks

7.

What does a transposed convolution layer do in a GAN generator?

a)

Reduces spatial resolution

b)

Acts as a pooling layer

c)

Upsamples feature maps

d)

Computes gradients

8.

In a Conditional GAN, how is the class label typically fed into the generator?

a)

As an output

b)

Ignored

c)

Concatenated with the input noise

d)

Used to initialize weights

9.

What is the key idea behind CycleGAN?

a)

Supervised learning from labeled pairs

b)

Learning mappings using conditional labels

c)

Unsupervised image-to-image translation using cycle consistency

d)

Reducing GAN parameters

10.

Pix2Pix is best described as:

a)

A style transfer GAN without supervision

b)

A paired image-to-image translation GAN

c)

A text-to-image generator

d)

A classifier based on adversarial loss

11.

What is one benefit of using Wasserstein loss in GANs?

a)

It reduces training time

b)

It eliminates the need for a discriminator

c)

It provides more stable gradients for training

d)

It increases mode collapse

12.

Which component of StyleGAN is primarily responsible for style control?

a)

Discriminator

b)

Generator output layer

c)

Mapping network

d)

Latent vector z

13.

What does the term 'scalability' in BigGAN mainly refer to?

a)

Reducing model parameters

b)

Increasing batch size and model capacity

c)

Shrinking dataset size

d)

Using transformers

14.

How does latent space interpolation work in GANs?

a)

By averaging discriminator outputs

b)

By linearly combining latent vectors

c)

By modifying loss values

d)

By reducing input noise

15.

Which GAN variant is most suited for unpaired image translation tasks?

a)

DCGAN

b)

CycleGAN

c)

Pix2Pix

d)

BigGAN

16.

What is the primary objective of a GAN's generator?

a)

Classify real vs. fake data

b)

Minimize reconstruction error

c)

Generate data that can fool the discriminator

d)

Encode data into latent space

17.

What does GAN stand for?

a)

Generalized Adversarial Network

b)

Generative Adversarial Network

c)

Generative Analysis Network

d)

General Artificial Network

18.

Which two components make up a GAN?

a)

Encoder and Decoder

b)

Generator and Discriminator

c)

Classifier and Regressor

d)

Generator and Encoder

19.

In a GAN, what type of learning is primarily used?

a)

Supervised Learning

b)

Unsupervised Learning

c)

Reinforcement Learning

d)

None of the above

20.

In a GAN, what type of learning is primarily used?

a)

Supervised Learning

b)

Unsupervised Learning

c)

Reinforcement Learning

d)

Self-supervised Learning

21.

What loss function is commonly used in GANs?

a)

Cross-entropy loss

b)

Mean squared error

c)

Hinge loss

d)

Triplet loss

22.

What is the primary objective of a GAN's generator?

a)

To classify images

b)

To detect anomalies

c)

To generate realistic data that can fool the discriminator

d)

To calculate loss functions

23.

Which of the following is NOT a role of the discriminator in a GAN?

a)

Distinguish real data from fake data

b)

Provide gradient feedback to the generator

c)

Generate new data samples

d)

Improve over time to detect better fakes

24.

What kind of loss is used in the original GAN paper for the generator?

a)

Mean Squared Error

b)

Binary Cross-Entropy

c)

Hinge Loss

d)

Wasserstein Loss

25.

What is 'mode collapse' in GANs?

a)

The generator learns the entire data distribution

b)

The discriminator becomes too weak

c)

The generator produces limited types of outputs

d)

The training fails due to overfitting

26.

Which of the following helps in improving GAN training stability?

a)

Decreasing the number of epochs

b)

Using ReLU in the discriminator

c)

One-sided label smoothing

d)

Removing the discriminator

27.

What architecture does DCGAN use in place of fully connected layers?

a)

RNN layers

b)

LSTM layers

c)

Convolutional layers

d)

Transformer blocks

28.

What does a transposed convolution layer do in a GAN generator?

a)

Reduces spatial resolution

b)

Acts as a pooling layer

c)

Upsamples feature maps

d)

Computes gradients

29.

In a Conditional GAN, how is the class label typically fed into the generator?

a)

As an output

b)

Ignored

c)

Concatenated with the input noise

d)

Used to initialize weights

30.

What is the key idea behind CycleGAN?

a)

Supervised learning from labeled pairs

b)

Learning mappings using conditional labels

c)

Unsupervised image-to-image translation using cycle consistency

d)

Reducing GAN parameters

31.

Pix2Pix is best described as:

a)

A style transfer GAN without supervision

b)

A paired image-to-image translation GAN

c)

A text-to-image generator

d)

A classifier based on adversarial loss

32.

Which technique helps mitigate mode collapse?

a)

Batch Normalization

b)

Gradient Penalty

c)

Dropout

d)

Data Augmentation

33.

What is a common cause of instability in GAN training?

a)

Learning rate mismatch

b)

Overfitting of the discriminator

c)

Insufficient data

d)

All of the above

34.

Which architecture uses convolutional layers for image generation?

a)

DCGAN

b)

CycleGAN

c)

StyleGAN

d)

All of the above

35.

Which GAN variant is known for high-resolution image synthesis?

a)

BigGAN

b)

CycleGAN

c)

Pix2Pix

d)

StyleGAN

36.

What is the main advantage of StyleGAN over traditional GANs?

a)

Faster training

b)

Better image resolution

c)

Improved control over generated features

d)

Simpler architecture

37.

Which task is CycleGAN particularly suited for?

a)

Image classification

b)

Image-to-image translation without paired data

c)

Text generation

d)

Time-series forecasting

38.

What is the primary use case of Pix2Pix?

a)

Image super-resolution

b)

Image-to-image translation with paired data

c)

Text-to-image generation

d)

Anomaly detection

39.

In StyleGAN, what does the mapping network do?

a)

Maps latent vectors to styles

b)

Generates images

c)

Classifies images

d)

None of the above

40.

What is latent space manipulation in GANs?

a)

Modifying the training data

b)

Altering the generator's weights

c)

Interpolating between latent vectors to generate new images

d)

Adjusting the discriminator's parameters

41.

What is the primary advantage of using Deep Convolutional GANs (DCGANs) over traditional GANs?

a)

Faster training convergence

b)

Improved image resolution

c)

Enhanced stability during training

d)

Ability to generate 3D images

42.

In CycleGAN, what is the purpose of the cycle consistency loss?

a)

To ensure the generator learns to produce diverse outputs

b)

To prevent mode collapse

c)

To ensure that an image translated to another domain can be translated back to the original domain

d)

To penalize the discriminator for making incorrect classifications

43.

Which of the following is a key feature of StyleGAN's generator architecture?

a)

Use of residual blocks for feature refinement

b)

Progressive growing of layers during training

c)

Mapping of latent vectors to an intermediate latent space before image generation

d)

Incorporation of attention mechanisms for feature selection

44.

Which of the following techniques is commonly used to address mode collapse in GANs?

a)

Batch normalization

b)

Two time-scale update rule (TTUR)

c)

Data augmentation

d)

Early stopping

45.

Which loss function is commonly used in Wasserstein GANs (WGANs) to improve training stability?

a)

Binary cross-entropy loss

b)

Mean squared error loss

c)

Earth Mover's Distance (Wasserstein loss)

d)

Hinge loss

46.

Which metric is commonly used to evaluate the quality of images generated by GANs?

a)

Mean squared error (MSE)

b)

Fréchet Inception Distance (FID)

c)

Precision and recall

d)

Cross-entropy loss

47.

What is the primary difference between a standard GAN and a Conditional GAN (cGAN)?

4 lines
48.

What is the primary difference between a standard GAN and a Conditional GAN (cGAN)?

a)

cGANs use a different loss function

b)

cGANs incorporate additional information (like labels) into both the generator and discriminator

c)

cGANs do not use a discriminator

d)

cGANs generate images without any input noise

49.

Which of the following is a characteristic of BigGAN?

a)

It uses a smaller network architecture for faster training

b)

It is designed for high-resolution image generation with improved scalability

c)

It operates on a single GPU for training

d)

It focuses on generating text data rather than images

50.

Which of the following is a key difference between CycleGAN and Pix2Pix?

a)

CycleGAN requires paired images for training, while Pix2Pix does not

b)

CycleGAN uses a cycle consistency loss to ensure bidirectional translation, while Pix2Pix uses paired images for supervised learning

c)

CycleGAN is used for text generation, while Pix2Pix is used for image generation

d)

CycleGAN uses a single generator, while Pix2Pix uses multiple generators

51.

Which GAN variant introduces a Wasserstein loss to improve training stability?

a)

CGAN

b)

WGAN

c)

DCGAN

d)

CycleGAN

52.

In Conditional GANs (cGANs), what additional information is fed to both the generator and discriminator?

a)

Noise

b)

Labels

c)

Latent vectors

d)

Loss values

53.

What innovation does CycleGAN introduce to perform image-to-image translation without paired data?

4 lines
54.

True or False: The Wasserstein distance is more stable than JS-divergence in the context of GANs.

4 lines
55.

What is the purpose of using gradient penalty in WGAN-GP?

4 lines
56.

Which GAN variant is best suited for high-resolution image synthesis?

a)

CycleGAN

b)

WGAN

c)

Progressive Growing GAN

d)

InfoGAN

57.

InfoGAN is designed to:

a)

Translate images from one domain to another

b)

Learn disentangled and interpretable latent representations

c)

Improve training stability

d)

Enhance discriminator performance

58.

Which loss function does StyleGAN use to improve perceptual quality?

a)

MSE Loss

b)

Hinge Loss

c)

Perceptual Loss

d)

Fréchet Inception Distance (FID)

59.

In WGAN, what key mathematical property is enforced to ensure the Wasserstein distance is valid?

a)

Symmetry of the generator output

b)

Lipschitz continuity of the discriminator

c)

Orthogonality of latent vectors

d)

Convexity of the loss function

60.

Which technique is used in WGAN-GP to enforce the 1-Lipschitz constraint without weight clipping?

a)

Spectral Normalization

b)

Gradient Penalty

c)

Batch Normalization

d)

Dropout Regularization

61.

What is a common symptom of mode collapse in GAN training?

a)

Discriminator outputs near 0.5 for all inputs

b)

Generator produces diverse but inaccurate samples

c)

Generator produces identical or very similar outputs

d)

Training loss of the generator increases steadily

62.

The Fréchet Inception Distance (FID) is used to evaluate GANs by comparing:

a)

Pixel-wise distances between real and generated images

b)

Log-likelihood of real vs generated data

c)

Feature distributions of real and generated images in a pretrained network

d)

Number of distinct modes captured by the generator

63.

Which of the following GAN models is best known for high-quality face generation and style transfer between image features?

a)

CycleGAN

b)

WGAN

c)

StyleGAN2

d)

InfoGAN

64.

In StyleGAN, the intermediate latent space W is introduced to:

a)

Increase randomness of the generated samples

b)

Improve disentanglement of semantic features

c)

Reduce training time

d)

Enforce mode collapse

65.

What is the role of the mapping network in StyleGAN?

a)

Projects images to labels

b)

Converts random noise z to intermediate latent code w

c)

Calculates perceptual loss

d)

Discriminates real vs fake images

66.

In Progressive Growing of GANs (PGGAN), what is the primary advantage of growing the network over time?

a)

Lower computation cost during inference

b)

Reduced number of training epochs

c)

Stability and quality improvement for high-resolution image generation

d)

Elimination of discriminator collapse

67.

Which GAN model introduces cycle-consistency loss to allow translation between domains without paired data?

a)

Pix2Pix

b)

CycleGAN

c)

BigGAN

d)

DCGAN

68.

The discriminator in a GAN may become too powerful during training, leading to:

a)

Mode expansion

b)

Generator gradient vanishing

c)

Faster convergence

d)

Improved diversity

69.

Spectral Normalization in GANs is used to:

a)

Enforce sparsity in weights

b)

Limit the Lipschitz constant of the discriminator

c)

Reduce mode collapse

d)

Normalize generator output

70.

Which of the following loss functions is used in Hinge GANs?

a)

Binary Cross Entropy

b)

L2 Loss

c)

Hinge Loss

d)

Wasserstein Loss

71.

In BigGAN, which of the following is a major enhancement for high-quality image generation?

a)

Unpaired image translation

b)

Conditional Batch Normalization

c)

Cycle-consistency loss

d)

Transformer blocks

72.

One of the goals of InfoGAN is to:

a)

Reduce training time

b)

Improve adversarial loss

c)

Learn interpretable and disentangled latent variables

d)

Transfer image style between two domains

73.

Why is training GANs considered a non-convex optimization problem?

a)

It uses dropout in both networks

b)

The objective function is linear in the generator

c)

The loss surface has multiple saddle points and local minima

d)

Discriminator is always convex

74.

Which of the following techniques is commonly used to stabilize GAN training?

a)

Dropout

b)

Batch Normalization

c)

Label Smoothing

d)

Both b and c

75.

What distinguishes GANformer from traditional GANs?

a)

Use of transformers for hierarchical generation

b)

Progressive upscaling

c)

Mutual information maximization

d)

Graph-based latent encoding