ISYE 7406 Week 4 Quiz Key Questions & Answers
1. Supervised vs. Unsupervised Learning
Question: Distinguish between supervised and unsupervised learning.
Answer:
Supervised learning uses labeled data (input-output pairs) to train models.
Unsupervised learning uses unlabeled data to find hidden patterns or groupings.
Supervised learning vs. Unsupervised learning distinction.
2. Assumptions in LDA and QDA
Question: What assumptions do LDA (Linear Discriminant Analysis) and QDA
(Quadratic Discriminant Analysis) make?
Answer:
Both assume the predictors follow a multivariate normal distribution.
LDA assumes equal covariance matrices across classes, while QDA allows
different covariance matrices.
LDA and QDA assume multivariate normality.
3. Independence in Gaussian Components
Question: What does the Gaussian assumption imply about predictors?
Answer:
It assumes each component of is an independent Gaussian (normally
distributed) variable.
This simplifies modeling but may not hold in practice.
Independent Gaussian components assumption.
4. Parametric vs. Nonparametric Methods
Question: Are LDA and QDA parametric or nonparametric?
Answer:
They are parametric methods because they rely on distributional assumptions
(normality, covariance).
Nonparametric methods (like k-NN) do not assume a specific distribution.
LDA/QDA are parametric.
5. Classification Boundaries
Question: How do LDA and QDA differ in classification boundaries?
Answer:
LDA: Linear boundaries between classes.
QDA: Quadratic boundaries, more flexible but requires more data.
LDA = linear, QDA = quadratic.
6. Practical Applications
Question: When should you use LDA vs. QDA?
Answer:
LDA: Better when sample size is small and covariance matrices are similar.
, QDA: Better when covariance structures differ significantly and enough data is
available.
7. Model Evaluation
Question: How do you evaluate supervised learning models?
Answer:
Use metrics like accuracy, precision, recall, F1-score, ROC curves.
Cross-validation is often applied to avoid overfitting.
Q7. Which assumption differentiates LDA from QDA?
A. Equal covariance matrices across classes
B. Nonparametric estimation
C. Independence of predictors
D. Quadratic decision boundaries
Q8. What type of boundaries does QDA produce?
A. Linear
B. Exponential
C. Circular
D. Quadratic
Q9. Which learning method uses labeled data?
A. Unsupervised learning
B. Reinforcement learning
C. Supervised learning
D. Clustering
Q10. Which distributional assumption underlies LDA and QDA?
A. Exponential distribution
B. Poisson distribution
C. Uniform distribution
D. Multivariate normal distribution
Q11. Are LDA and QDA parametric or nonparametric?
A. Nonparametric
B. Semi-parametric
C. Parametric
D. Bayesian
Q12. Which method is more data-efficient when sample sizes are small?
A. k-NN
B. QDA
C. LDA
D. Decision Trees
Q13. Which evaluation metric balances precision and recall?
1. Supervised vs. Unsupervised Learning
Question: Distinguish between supervised and unsupervised learning.
Answer:
Supervised learning uses labeled data (input-output pairs) to train models.
Unsupervised learning uses unlabeled data to find hidden patterns or groupings.
Supervised learning vs. Unsupervised learning distinction.
2. Assumptions in LDA and QDA
Question: What assumptions do LDA (Linear Discriminant Analysis) and QDA
(Quadratic Discriminant Analysis) make?
Answer:
Both assume the predictors follow a multivariate normal distribution.
LDA assumes equal covariance matrices across classes, while QDA allows
different covariance matrices.
LDA and QDA assume multivariate normality.
3. Independence in Gaussian Components
Question: What does the Gaussian assumption imply about predictors?
Answer:
It assumes each component of is an independent Gaussian (normally
distributed) variable.
This simplifies modeling but may not hold in practice.
Independent Gaussian components assumption.
4. Parametric vs. Nonparametric Methods
Question: Are LDA and QDA parametric or nonparametric?
Answer:
They are parametric methods because they rely on distributional assumptions
(normality, covariance).
Nonparametric methods (like k-NN) do not assume a specific distribution.
LDA/QDA are parametric.
5. Classification Boundaries
Question: How do LDA and QDA differ in classification boundaries?
Answer:
LDA: Linear boundaries between classes.
QDA: Quadratic boundaries, more flexible but requires more data.
LDA = linear, QDA = quadratic.
6. Practical Applications
Question: When should you use LDA vs. QDA?
Answer:
LDA: Better when sample size is small and covariance matrices are similar.
, QDA: Better when covariance structures differ significantly and enough data is
available.
7. Model Evaluation
Question: How do you evaluate supervised learning models?
Answer:
Use metrics like accuracy, precision, recall, F1-score, ROC curves.
Cross-validation is often applied to avoid overfitting.
Q7. Which assumption differentiates LDA from QDA?
A. Equal covariance matrices across classes
B. Nonparametric estimation
C. Independence of predictors
D. Quadratic decision boundaries
Q8. What type of boundaries does QDA produce?
A. Linear
B. Exponential
C. Circular
D. Quadratic
Q9. Which learning method uses labeled data?
A. Unsupervised learning
B. Reinforcement learning
C. Supervised learning
D. Clustering
Q10. Which distributional assumption underlies LDA and QDA?
A. Exponential distribution
B. Poisson distribution
C. Uniform distribution
D. Multivariate normal distribution
Q11. Are LDA and QDA parametric or nonparametric?
A. Nonparametric
B. Semi-parametric
C. Parametric
D. Bayesian
Q12. Which method is more data-efficient when sample sizes are small?
A. k-NN
B. QDA
C. LDA
D. Decision Trees
Q13. Which evaluation metric balances precision and recall?
