CS 7643 Quiz 4 Verified Multiple Choice and
Conceptual Actual Frequently Tested Exam
Questions With Reviewed 100% Correct
Detailed Answers
Guaranteed Pass!!Current Update!!
Q1. What is the purpose of embeddings in deep learning?
A. To store full datasets as tables
B. To encode entities into vector spaces capturing similarity
C. To replace convolutional layers
D. To reduce network size
Q2. . Which gate in LSTM decides what information to forget?
A. Input gate
B. Forget gate
C. Output gate
D. Candidate update
Q3. The LSTM cell state is updated using:
A. Forget + input updates
B. Only output gate
C. Bias parameters
D. Random noise
Q4. In graph embeddings, which nodes are expected to have more similar
embeddings?
A. Randomly selected nodes
B. Unconnected nodes
C. Connected nodes
D. All nodes in the same layer
,Q5. Graph embeddings are often trained using:
A. Fully supervised labels
B. Random initialization only
C. Objectives where neighbors have similar embeddings
D. Dropout-based regularization
Q6. Why are embeddings useful for downstream tasks?
A. They eliminate the need for training
B. They provide dense, low-dimensional features
C. They replace backpropagation
D. They guarantee zero error
Q7. Which is an example of embedding use in NLP?
A. Word2Vec
B. Gantt chart
C. Support Vector Machine
D. PCA
Q8. What does an embedding vector typically encode?
A. File sizes
B. Semantic similarity
C. Random noise
D. Loss function gradients
Q9. Which is not a property of good embeddings?
A. Capture contextual similarity
B. Can generalize to new tasks
C. Are always interpretable
D. Compact representation
Q10. Which of the following tasks benefits most from graph embeddings?
A. Node classification
B. Sorting algorithms
C. Matrix inversion
D. Arithmetic operations
, Q11. Embedding learning is often:
A. Reinforcement-based only
B. Fully unsupervised or self-supervised
C. Dependent only on CNNs
D. Done by human labeling
Q12. What is the dimensionality of embeddings compared to original features?
A. Always larger
B. Always smaller
C. Usually lower, more compact
D. Must match original
Q13. Why are MLPs inadequate for NLP tasks?
A. Too accurate
B. Cannot easily handle variable-length sequences
C. Lack gradient descent
D. Cannot use GPUs
Q14. Which structure is not inherently modeled by MLPs?
A. Temporal dependencies
B. Linear regression
C. Nonlinear mappings
D. Hidden layers
Q3. Why does MLP size grow with input sequence length?
A. They require one weight per possible input position
B. They use dynamic memory
C. They are convolutional by design
D. They require recurrent feedback
Q4. MLPs do not provide a mechanism for:
A. Addition
B. Holding state over time
C. Matrix multiplication
D. Classification
Conceptual Actual Frequently Tested Exam
Questions With Reviewed 100% Correct
Detailed Answers
Guaranteed Pass!!Current Update!!
Q1. What is the purpose of embeddings in deep learning?
A. To store full datasets as tables
B. To encode entities into vector spaces capturing similarity
C. To replace convolutional layers
D. To reduce network size
Q2. . Which gate in LSTM decides what information to forget?
A. Input gate
B. Forget gate
C. Output gate
D. Candidate update
Q3. The LSTM cell state is updated using:
A. Forget + input updates
B. Only output gate
C. Bias parameters
D. Random noise
Q4. In graph embeddings, which nodes are expected to have more similar
embeddings?
A. Randomly selected nodes
B. Unconnected nodes
C. Connected nodes
D. All nodes in the same layer
,Q5. Graph embeddings are often trained using:
A. Fully supervised labels
B. Random initialization only
C. Objectives where neighbors have similar embeddings
D. Dropout-based regularization
Q6. Why are embeddings useful for downstream tasks?
A. They eliminate the need for training
B. They provide dense, low-dimensional features
C. They replace backpropagation
D. They guarantee zero error
Q7. Which is an example of embedding use in NLP?
A. Word2Vec
B. Gantt chart
C. Support Vector Machine
D. PCA
Q8. What does an embedding vector typically encode?
A. File sizes
B. Semantic similarity
C. Random noise
D. Loss function gradients
Q9. Which is not a property of good embeddings?
A. Capture contextual similarity
B. Can generalize to new tasks
C. Are always interpretable
D. Compact representation
Q10. Which of the following tasks benefits most from graph embeddings?
A. Node classification
B. Sorting algorithms
C. Matrix inversion
D. Arithmetic operations
, Q11. Embedding learning is often:
A. Reinforcement-based only
B. Fully unsupervised or self-supervised
C. Dependent only on CNNs
D. Done by human labeling
Q12. What is the dimensionality of embeddings compared to original features?
A. Always larger
B. Always smaller
C. Usually lower, more compact
D. Must match original
Q13. Why are MLPs inadequate for NLP tasks?
A. Too accurate
B. Cannot easily handle variable-length sequences
C. Lack gradient descent
D. Cannot use GPUs
Q14. Which structure is not inherently modeled by MLPs?
A. Temporal dependencies
B. Linear regression
C. Nonlinear mappings
D. Hidden layers
Q3. Why does MLP size grow with input sequence length?
A. They require one weight per possible input position
B. They use dynamic memory
C. They are convolutional by design
D. They require recurrent feedback
Q4. MLPs do not provide a mechanism for:
A. Addition
B. Holding state over time
C. Matrix multiplication
D. Classification
