,
, CS 7643 Quiz 4 – Concepts Notes | 2026
Update | with Complete Solutions.
🧠 SECTION 1: EMBEDDINGS & REPRESENTATION
LEARNING (Questions 1–6)
Question 1
What is the purpose of embeddings in deep learning?
A) To store full datasets as tables for relational queries
B) To encode entities into vector spaces where similarity reflects semantic or
relational closeness
C) To replace convolutional layers with fully connected layers
D) To reduce the number of training epochs required
E) To convert categorical labels into one-hot vectors only
Correct Answer: B
Rationale: Embeddings are learned mappings from discrete entities (words,
nodes, users, items) to continuous vector spaces. The key property is that
entities with similar meanings or relationships end up with similar vectors
(e.g., word2vec: "king" - "man" + "woman" ≈ "queen") . They convert
sparse, high-dimensional representations into dense, low-dimensional ones
while preserving semantic structure.
Question 2
Graph Embedding optimizes an objective that:
, CS 7643 Quiz 4 – Concepts Notes | 2026
Update | with Complete Solutions.
🧠 SECTION 1: EMBEDDINGS & REPRESENTATION
LEARNING (Questions 1–6)
Question 1
What is the purpose of embeddings in deep learning?
A) To store full datasets as tables for relational queries
B) To encode entities into vector spaces where similarity reflects semantic or
relational closeness
C) To replace convolutional layers with fully connected layers
D) To reduce the number of training epochs required
E) To convert categorical labels into one-hot vectors only
Correct Answer: B
Rationale: Embeddings are learned mappings from discrete entities (words,
nodes, users, items) to continuous vector spaces. The key property is that
entities with similar meanings or relationships end up with similar vectors
(e.g., word2vec: "king" - "man" + "woman" ≈ "queen") . They convert
sparse, high-dimensional representations into dense, low-dimensional ones
while preserving semantic structure.
Question 2
Graph Embedding optimizes an objective that:
