# Deep Learning | Week 12

Course Name: Deep Learning

#### Q1. When the dice coefficient between two samples will be one?a. When there is a perfect overlap between the two samples.b. When there is no overlap between the two samples.c. It cannot be one.d. If the inner product of two samples is one.

Answer: a. When there is a perfect overlap between the two samples.

Q2. There are two distributions. The first distribution, P is a uniform distribution between [-1, 1]. Another distribution, Q is a Normal distribution. What will be the KL(Q| |P)?
a. 0.5
b. 0.0
c. 10
d. infinity

These are NPTEL Deep Learning Week 12 Assignment 12 Answers

Q3. Which of the following is True regarding the reconstruction loss (realized as mean squared error between input and predicted signal) of standard auto-encoder?
a. Such loss is not differentiable and cannot be used for back propagation
b. Such loss tends to form distinct clusters in latent space
c. Such loss cannot be optimized with gradient descent
d. None of the above

Answer: b. Such loss tends to form distinct clusters in latent space

Q4. For an auto-encoder, suppose we give an input signal, x and reconstruct a signal y. Which one of the following objective functions can be MINIMIZED to train the parameters of the auto encoder using gradient descent optimizer?
a. L(x, y) = exp -(lx-yl)
b. L(x,y)= -log(lx-yl)
c. L(x, y) = exp(|x-yl)
d. L(x,y) = (x + y)²

Answer: c. L(x, y) = exp(|x-yl)

Q5. Suppose we have a 2N dimensional Normal distribution in which we assume all components are independent of each other. What will be the size (number of elements) of the vector to fully represent the covariance matrix of this distribution?
a. N
b. 2N
c N/2
d. N/4

These are NPTEL Deep Learning Week 12 Assignment 12 Answers

Q6. What will happen if we do not enforce KL divergence loss in VAE latent code space?
a. The latent code distribution will be mimic zero mean and unit variance Normal distribution
b. Network will learn to form distinctive clusters with high standard deviation for each cluster
c. Network will learn to form distinctive clusters with low standard deviation for each cluster
d. None of the above

Answer: c. Network will learn to form distinctive clusters with low standard deviation for each cluster

These are NPTEL Deep Learning Week 12 Assignment 12 Answers

Q7. Figure shows latent vector addition of two concepts of “man without a hat” and “hat”. What is
expected from the resultant vector?

a. Hat without man
b. Man with hat
c. Woman with hat
d. Woman without hat

These are NPTEL Deep Learning Week 12 Assignment 12 Answers

Q8. Which one of the following statements is True in the original GAN training?
a. It is desired that the Discriminator loss monotonically goes down
b. It is desired that the Generator loss monotonically goes down
c. It is desired that the Discriminator loss monotonically goes down while the Discriminator loss monotonically goes up
d. It is desired that neither of the losses of Discriminator or Generator monotonically goes up or down monotonically

Answer: d. It is desired that neither of the losses of Discriminator or Generator monotonically goes up or down monotonically

These are NPTEL Deep Learning Week 12 Assignment 12 Answers

Q9. Which one of the following statements is true about Variational Autoencoder (VAE)?
a. VAE can only be applied on monochrome images
b. VAE reconstructions tend to be blurry
c. VAE reconstructions always have high frequency preserving details
d. VAE latent space is designed to be NOT smooth

Answer: b. VAE reconstructions tend to be blurry

These are NPTEL Deep Learning Week 12 Assignment 12 Answers

Q10. Suppose we have trained a Variational Auto-encoder (VAE) on faces with 4D latent code and after convergence, the mean vector and standard deviation vector is given by [2.1, -1.9, 3.8, 0.9] and [0.5, 0.8, 0.6, 0.2] respectively. Now, in evaluation phase, we pass a new face image (belonging to the same domain as those used in training phase). Which of the following latent code is most probable to be encountered when we pass this new face image?
a. [1.0,3.9,2.9,3.2]
b. [2.0,-1.92,3.8,0.92]
c [2.1,23,-39,10]
d. [3.2,-3.6,09,12]