Wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

Autoencoders and Generative Models Quiz

Total questions: 28

Worksheet time: 14mins

Name
Class
Date
1.

What is the primary purpose of an autoencoder?

a)

Classification

b)

Data compression and reconstruction

c)

Supervised learning

d)

Label prediction

2.

Autoencoders are trained using which type of learning?

a)

Supervised

b)

Reinforcement

c)

Unsupervised

d)

Transfer

3.

The compressed representation learned by an autoencoder is called:

a)

Feature map

b)

Hidden layer

c)

Latent space

d)

Output space

4.

Which component of an autoencoder performs dimensionality reduction?

a)

Decoder

b)

Bottleneck

c)

Encoder

d)

Loss function

5.

The bottleneck layer mainly helps to:

a)

Increase accuracy

b)

Prevent memorization

c)

Increase model size

d)

Speed up training

6.

Latent variables are best described as:

a)

Observable outputs

b)

Noise variables

c)

Underlying hidden variables

d)

Labels

7.

Which of the following is NOT a common application of autoencoders?

a)

Image denoising

b)

Anomaly detection

c)

Facial recognition

d)

Rule-based reasoning

8.

Dimensionality reduction refers to:

a)

Increasing features

b)

Removing labels

c)

Compressing data into fewer dimensions

d)

Adding noise

9.

A 28×28 grayscale MNIST image is represented as how many dimensions?

a)

256

b)

512

c)

784

d)

1024

10.

Why is dimensionality reduction useful in ML?

a)

Increases noise

b)

Reduces accuracy

c)

Improves efficiency and performance

d)

Removes patterns

11.

Which component reconstructs the original input?

a)

Encoder

b)

Bottleneck

c)

Decoder

d)

Optimizer

12.

In a typical encoder, the number of neurons in each layer:

a)

Increases

b)

Remains constant

c)

Decreases

d)

Is random

13.

The decoder architecture generally:

a)

Mirrors the encoder

b)

Has fewer layers

c)

Uses no activation

d)

Has constant size

14.

VAEs differ from standard autoencoders because they are:

a)

Deterministic

b)

Supervised

c)

Probabilistic

d)

Rule-based

15.

VAEs encode latent variables as:

a)

Single values

b)

Discrete labels

c)

Probability distributions

d)

Binary codes

16.

In VAEs, latent variables are represented using:

a)

Mean only

b)

Variance only

c)

Mean and standard deviation

d)

Weights and bias

17.

Reconstruction loss measures:

a)

Model complexity

b)

Distance between input and output

c)

Training speed

d)

Feature importance

18.

KL divergence measures the difference between:

a)

Two datasets

b)

Two loss functions

c)

Two probability distributions

d)

Two neural networks

19.

Generative AI primarily focuses on:

a)

Classification

b)

Prediction

c)

Content creation

d)

Rule execution

20.

Which of the following is NOT a generative model?

a)

VAE

b)

GAN

c)

Logistic Regression

d)

Diffusion Model

21.

Generative models learn:

a)

Decision boundaries

b)

Data distribution

c)

Feature labels

d)

Class margins

22.

Discriminative models mainly learn:

a)

Joint probability

b)

Marginal probability

c)

Conditional probability

d)

Latent space

23.

The core assumption of Naive Bayes is:

a)

Features are dependent

b)

Features are independent

c)

Labels are continuous

d)

Data is sequential

24.

Maximum Likelihood Estimation (MLE) aims to:

a)

Minimize error

b)

Maximize accuracy

c)

Maximize likelihood of observed data

d)

Reduce parameters

25.

Additive (Laplace) smoothing is used to:

a)

Reduce noise

b)

Increase variance

c)

Avoid zero probabilities

d)

Remove bias

26.

Representation learning refers to:

a)

Manual feature design

b)

Learning rules

c)

Automatically learning meaningful features

d)

Data labeling

27.

What is the main advantage of using a Variational Autoencoder (VAE) over a traditional autoencoder?

a)

Faster training

b)

Better reconstruction quality

c)

Probabilistic interpretation of latent space

d)

More complex architecture

28.

In the context of neural networks, what does the term 'overfitting' refer to?

a)

Model performs well on training data but poorly on unseen data

b)

Model has too many parameters

c)

Model is too simple to capture the underlying patterns

d)

Model is unable to learn from the training data