Exam NCA-GENM Topic 1 Question 386 Discussion

Actual exam question for NVIDIA's NCA-GENM exam
Question #: 386
Topic #: 1
You are experimenting with different loss functions for training a Variational Autoencoder (VAE) to generate images. You observe that using only the reconstruction loss (e.g., Mean Squared Error) results in blurry images. What other loss component is typically added to the VAE objective function to encourage the latent space to be well-structured and generate sharper images?

Suggested Answer: B Vote an answer

The Kullback-Leibler (KL) divergence loss is a crucial component of the VAE objective function. It measures the difference between the learned latent space distribution and a prior distribution (typically a standard Gaussian). Adding the KL divergence loss encourages the latent space to be well-structured and continuous, which helps generate sharper and more realistic images. The other loss functions serve different purposes and are not typically used in VAEs for this specific reason. Cross-entropy is for classification. Perceptual loss helps in transferring styles. Contrastive loss used to learn embedding. Hinge loss mostly used in SVM.

by Hazel at Sep 18, 2025, 01:35 AM

Limited Time Offer

15%

Off

Get Premium NCA-GENM Questions as Interactive Self Test Engine or PDF

Comments

Chosen Answer:
This is a voting comment (?) , you can switch to a simple comment.
Switch to a voting comment New
Nick name: Submit Cancel
A voting comment increases the vote count for the chosen answer by one.

Upvoting a comment with a selected answer will also increase the vote count towards that answer by one. So if you see a comment that you already agree with, you can upvote it instead of posting a new comment.
0
0
0
10