CS7643 Verified Multiple Choice and
Conceptual Actual Frequently Tested Exam
Questions With Reviewed 100% Correct
Detailed Answers
Guaranteed Pass!!Current Update!!
Q1. Which of the following are common features extracted by convolutional filters
in early layers of CNNs?
A. Sentiment, syntax, grammar, embeddings
B. Edges, colors, textures, motifs
C. Hidden states, recurrent steps, embeddings
D. Rewards, policies, value functions
Answer: B
Q2. What is a receptive field in CNNs?
A. The full dataset size the network is trained on.
B. A region of an image patch from which a node receives input.
C. The number of kernels applied per channel.
D. The loss function used in convolutional training.
Answer: B
Q3. What is the key difference between convolution and cross-correlation in
image processing?
A. Convolution sums pixel values, cross-correlation averages them.
, B. Convolution flips the kernel before dot product; cross-correlation does not.
C. Cross-correlation is only used for color channels.
D. Convolution requires padding while cross-correlation does not.
Answer: B
Q4. Why is using an image patch advantageous?
A. It eliminates the need for kernels.
B. It reduces input parameters and maintains spatial information.
C. It avoids weight sharing.
D. It guarantees rotational invariance.
Answer: B
Q5. What does weight sharing in CNNs imply?
A. Different kernels must share the same weights.
B. A kernel uses the same weights for all patches in the image.
C. Each channel shares weights with other channels.
D. Kernels are updated only once per epoch.
Answer: B
Q6. How many parameters are required for a single kernel operating on one
channel?
A. K1×K2K_1 \times K_2K1×K2
B. K1×K2+1K_1 \times K_2 + 1K1×K2+1
C. N×(K1+K2+1)N \times (K_1 + K_2 + 1)N×(K1+K2+1)
D. (K1×K2+1)×M(K_1 \times K_2 + 1) \times M(K1×K2+1)×M
Answer: B
Q7. With multiple feature extractions (M kernels), what is the total number of
parameters per layer?
Conceptual Actual Frequently Tested Exam
Questions With Reviewed 100% Correct
Detailed Answers
Guaranteed Pass!!Current Update!!
Q1. Which of the following are common features extracted by convolutional filters
in early layers of CNNs?
A. Sentiment, syntax, grammar, embeddings
B. Edges, colors, textures, motifs
C. Hidden states, recurrent steps, embeddings
D. Rewards, policies, value functions
Answer: B
Q2. What is a receptive field in CNNs?
A. The full dataset size the network is trained on.
B. A region of an image patch from which a node receives input.
C. The number of kernels applied per channel.
D. The loss function used in convolutional training.
Answer: B
Q3. What is the key difference between convolution and cross-correlation in
image processing?
A. Convolution sums pixel values, cross-correlation averages them.
, B. Convolution flips the kernel before dot product; cross-correlation does not.
C. Cross-correlation is only used for color channels.
D. Convolution requires padding while cross-correlation does not.
Answer: B
Q4. Why is using an image patch advantageous?
A. It eliminates the need for kernels.
B. It reduces input parameters and maintains spatial information.
C. It avoids weight sharing.
D. It guarantees rotational invariance.
Answer: B
Q5. What does weight sharing in CNNs imply?
A. Different kernels must share the same weights.
B. A kernel uses the same weights for all patches in the image.
C. Each channel shares weights with other channels.
D. Kernels are updated only once per epoch.
Answer: B
Q6. How many parameters are required for a single kernel operating on one
channel?
A. K1×K2K_1 \times K_2K1×K2
B. K1×K2+1K_1 \times K_2 + 1K1×K2+1
C. N×(K1+K2+1)N \times (K_1 + K_2 + 1)N×(K1+K2+1)
D. (K1×K2+1)×M(K_1 \times K_2 + 1) \times M(K1×K2+1)×M
Answer: B
Q7. With multiple feature extractions (M kernels), what is the total number of
parameters per layer?
