Variational Autoencoder in PyTorch - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of VAE latent variable sampling code

What is the shape of the sampled latent variable z after running this PyTorch code snippet?

PyTorch

import torch
import torch.nn.functional as F

batch_size = 16
latent_dim = 10
mu = torch.zeros(batch_size, latent_dim)
logvar = torch.zeros(batch_size, latent_dim)
std = torch.exp(0.5 * logvar)
eps = torch.randn_like(std)
z = mu + eps * std
print(z.shape)

A torch.Size([16, 10])

B torch.Size([10, 16])

C torch.Size([10])

D torch.Size([16])

💡 Hint

Remember that mu and logvar have shape (batch_size, latent_dim), and sampling keeps that shape.

The sampled latent variable z has the same shape as mu and logvar, which is (batch_size, latent_dim). Here, batch_size=16 and latent_dim=10, so z.shape is (16, 10).

❓ Model Choice

intermediate

Choosing the correct VAE loss function components

Which of the following correctly describes the two main components of the Variational Autoencoder loss function?

A Cross-entropy loss between input and output, plus L2 regularization on weights

B Mean squared error between latent variables, plus entropy of output distribution

C Reconstruction loss measuring output-input difference, plus KL divergence between approximate posterior and prior

D Hinge loss on output labels, plus KL divergence between input and output

💡 Hint

Think about what VAE tries to reconstruct and what distribution it tries to regularize.

The VAE loss has two parts: a reconstruction loss that measures how well the decoder recreates the input, and a KL divergence term that regularizes the latent distribution to be close to a prior (usually standard normal).

❓ Hyperparameter

advanced

Effect of latent dimension size in VAE

What is the most likely effect of increasing the latent dimension size in a Variational Autoencoder model?

A The KL divergence term becomes zero regardless of data

B The model will always perform worse due to increased computational cost

C The reconstruction loss will increase because latent space is too large

D The model can capture more complex data features but risks overfitting and less regularization

💡 Hint

Think about how latent space size affects model capacity and regularization.

Increasing latent dimension allows the model to represent more complex features, but it may overfit and the KL divergence regularization becomes weaker, reducing the model's ability to generalize.

🔧 Debug

advanced

Identifying error in VAE reparameterization code

What error will this PyTorch code raise when running the reparameterization step in a VAE?

PyTorch

def reparameterize(mu, logvar):
    std = torch.exp(0.5 * logvar)
    eps = torch.randn_like(std)
    return mu + eps * std

mu = torch.zeros(4, 5)
logvar = torch.zeros(4, 5)
z = reparameterize(mu, logvar)

A No error, code runs correctly

B RuntimeError due to shape mismatch in multiplication

C TypeError because torch.exp cannot take logvar tensor

D ValueError because std is negative

💡 Hint

Check the formula for standard deviation from log variance in VAE reparameterization.

No runtime error is raised: all tensors have compatible shapes (4, 5), torch.exp works on tensors, and std is non-negative. The code runs correctly (B). Note that for proper VAE reparameterization, use std = torch.exp(0.5 * logvar).

❓ Metrics

expert

Interpreting VAE training metrics

During VAE training, you observe the reconstruction loss steadily decreases but the KL divergence term remains near zero. What is the most likely explanation?

A The model is perfectly regularized and has optimal latent distribution

B The model ignores the latent space and behaves like a standard autoencoder

C The KL divergence term is not implemented correctly and is always zero

D The reconstruction loss is not computed properly and is misleading

💡 Hint

Think about what a near-zero KL divergence means for latent variable distribution.

A near-zero KL divergence means the approximate posterior matches the prior exactly, so the latent variables carry no information about the input. This means the decoder ignores the latent space and the model acts like a standard autoencoder.

Practice

1. What is the main purpose of the encoder in a Variational Autoencoder (VAE)?

easy

A. To calculate the loss function

B. To produce the mean and variance for the latent space distribution

C. To generate new data samples from noise

D. To directly reconstruct the input data

Solution

1. Step 1: Understand the role of the encoder in VAE

   The encoder processes input data and outputs parameters of a probability distribution, typically mean and variance, representing the latent space.

2. Step 2: Differentiate encoder from decoder and loss

   The decoder reconstructs data, and loss measures reconstruction and latent space closeness. Encoder specifically outputs mean and variance.

3. Final Answer:

   To produce the mean and variance for the latent space distribution -> Option B

4. Quick Check:

   Encoder outputs mean and variance = D [OK]

Hint: Encoder outputs distribution parameters, not data directly [OK]

Common Mistakes:

• Confusing encoder with decoder
• Thinking encoder reconstructs data
• Assuming encoder calculates loss

2. Which PyTorch function is typically used to sample from the latent space in a VAE using the reparameterization trick?

easy

A. torch.randn()

B. torch.sigmoid()

C. torch.softmax()

D. torch.exp()

Solution

1. Step 1: Identify sampling method in VAE

   The reparameterization trick samples from a normal distribution using random noise, typically generated by torch.randn() in PyTorch.

2. Step 2: Understand other functions' roles

   torch.sigmoid() and torch.softmax() are activation functions, torch.exp() computes exponentials, none sample noise.

3. Final Answer:

   torch.randn() -> Option A

4. Quick Check:

   Sampling noise uses torch.randn() = A [OK]

Hint: Sampling noise uses torch.randn(), not activations [OK]

Common Mistakes:

• Using torch.sigmoid() to sample noise
• Confusing softmax with sampling
• Using torch.exp() incorrectly for sampling

3. Given the following PyTorch code snippet for the reparameterization trick in a VAE, what is the shape of the output z if mu and log_var have shape (batch_size, latent_dim)?

std = torch.exp(0.5 * log_var)
epsilon = torch.randn_like(std)
z = mu + epsilon * std

medium

A. (batch_size, latent_dim)

B. (latent_dim, batch_size)

C. (batch_size, 1)

D. (1, latent_dim)

Solution

1. Step 1: Analyze shapes of mu and log_var

   Both mu and log_var have shape (batch_size, latent_dim), so std and epsilon will have the same shape due to torch.exp and torch.randn_like.

2. Step 2: Calculate shape of z

   Since z = mu + epsilon * std, all operands have shape (batch_size, latent_dim), so z also has shape (batch_size, latent_dim).

3. Final Answer:

   (batch_size, latent_dim) -> Option A

4. Quick Check:

   Output shape matches input batch and latent dims = A [OK]

Hint: Output shape matches mu and log_var shapes [OK]

Common Mistakes:

• Swapping batch_size and latent_dim
• Assuming output is reduced to (batch_size, 1)
• Confusing shape broadcasting rules

4. Identify the error in this VAE loss function implementation in PyTorch:

def vae_loss(recon_x, x, mu, log_var):
    BCE = nn.functional.binary_cross_entropy(recon_x, x)
    KLD = -0.5 * torch.sum(1 + log_var - mu.pow(2) - log_var.exp())
    return BCE + KLD

medium

A. Incorrect sign in KLD calculation

B. Using torch.sum instead of torch.mean for KLD

C. Missing reduction='sum' in binary_cross_entropy causing scale mismatch

D. Not detaching mu and log_var before loss

Solution

1. Step 1: Check binary_cross_entropy usage

   By default, binary_cross_entropy returns mean loss per element. Summing KLD but not BCE causes scale mismatch.

2. Step 2: Fix BCE reduction

   Adding reduction='sum' to BCE matches scale with KLD, ensuring proper combined loss.

3. Final Answer:

   Missing reduction='sum' in binary_cross_entropy causing scale mismatch -> Option C

4. Quick Check:

   BCE needs reduction='sum' to match KLD scale = B [OK]

Hint: Match loss scales by using reduction='sum' in BCE [OK]

Common Mistakes:

• Ignoring BCE reduction parameter
• Changing KLD sign incorrectly
• Detaching mu/log_var unnecessarily

5. You want to train a VAE on images of size 28x28 pixels with a latent dimension of 20. Which of the following model architectures is most appropriate for the encoder part in PyTorch?

hard

A. A recurrent neural network processing images as sequences, outputting 20 values

B. A single linear layer flattening input and outputting 20 values directly

C. A fully connected network with no nonlinearities outputting 28x28 values

D. A convolutional network with Conv2d layers reducing spatial size, ending with linear layers outputting mean and log variance vectors of size 20

Solution

1. Step 1: Consider input data type and size

   Images of 28x28 pixels are spatial data; convolutional layers efficiently extract features and reduce spatial dimensions.

2. Step 2: Match output to latent dimension

   The encoder must output two vectors (mean and log variance) each of size 20, so linear layers after convolutions are suitable.

3. Step 3: Evaluate other options

   Single linear layer ignores spatial structure; RNNs are not typical for images; fully connected without nonlinearities won't learn complex features.

4. Final Answer:

   A convolutional network with Conv2d layers reducing spatial size, ending with linear layers outputting mean and log variance vectors of size 20 -> Option D

5. Quick Check:

   Conv layers + linear output for mean/log_var = C [OK]

Hint: Use conv layers for images, linear for latent params [OK]

Common Mistakes:

• Using only linear layers ignoring image structure
• Choosing RNNs for image encoding
• Omitting nonlinearities in encoder