ML2 GPT2 HARD-CHOKE

Discuss the challenges of using SVM with a very large number of observations.

• a) SVM is highly efficient with a large number of observations

• b) SVM struggles with computational intensity, especially with large datasets

• c) Large datasets have no impact on SVM performance

• d) SVM is insensitive to overfitting

Explain the relationship between the Total Probability Theorem and conditional probability.

• a) Total Probability Theorem calculates joint probability

• b) Total Probability Theorem calculates conditional probability

• c) Total Probability Theorem calculates the probability of an event based on several mutually exclusive scenarios

• d) Total Probability Theorem is unrelated to conditional probability

Explain the importance of assessing Gaussian distribution assumptions in LDA and QDA.

• a) Gaussian distribution is irrelevant in LDA and QDA

• b) Non-Gaussian distributions lead to better model performance

• c) Violations of Gaussian distribution assumptions can affect model performance

• d) Gaussian distribution assumptions are only crucial for SVM

Discuss the advantages and limitations of using Logistic Regression in comparison to Support Vector Machines.

• a) Logistic Regression is more robust against overfitting

• b) Logistic Regression struggles with large feature spaces and non-linear features without transformations

• c) SVM is less efficient with a very large number of observations

• d) Logistic Regression does not provide probability scores

How does the choice of kernel in SVM impact model training and prediction?

• a) Kernel choice has no impact on SVM

• b) Kernel choice influences model training and prediction, especially with complex datasets

• c) Linear kernel is always the best choice

• d) Kernel choice is only relevant for one-vs-one SVM

When is PCA more of an exploratory or preparatory technique rather than a definitive algorithm?

• a) PCA is always definitive

• b) PCA depends on the scale of the data

• c) PCA is definitive when variables are uncorrelated

• d) PCA is unsupervised and does not consider the response variable when summarizing variability

How does the flexibility of Quadratic Discriminant Analysis (QDA) impact its performance compared to Linear Discriminant Analysis (LDA)?

• a) QDA has lower variance due to flexibility

• b) QDA introduces more parameters and can lead to higher variance

• c) LDA is more flexible than QDA

• d) Flexibility has no impact on performance

Discuss the drawbacks of using the One-vs-Rest approach in multi-class SVM.

• a) Provides consistent results

• b) Leads to imbalanced training sets

• c) Works well for small datasets

• d) Minimizes overfitting

Explain the role of eigenvalues and eigenvectors in Principal Component Analysis (PCA).

• a) Eigenvectors indicate the amount of variance captured

• b) Eigenvalues provide the direction of axes in the new feature space

• c) Eigenvectors ensure that the principal components are correlated

• d) Eigenvalues indicate the amount of variance captured by each principal component

Discuss the trade-offs between a large value and a small value of C in SVM.

• a) Large C leads to a larger margin and less overfitting

• b) Small C allows more misclassifications and might underfit

• c) Both large and small C result in overfitting

• d) C has no impact on SVM performance