2018 A1

Question 1

Write down the whitened data in terms of the SVD components above

Answer 1

Whitened data = V^T

Question 2

Contrast LDA and PCA to the information required for each data point in the data set.

Answer 2

PCA only requires the feature or characteristic information of a data point. LDA requires both the feature and the class information.

Question 3

Contrast LDA and PCA to the maximum number of features that can sensibly be obtained

Answer 3

LDA - k-1, with k being the number of classes.
PCA - d-1 where min(n,d). So that the sensible maximum would be n should n be smaller than d.

Question 4

Give the name of the classifier which is based on the fundamental assumption that the dimensions of x are conditionally independent.

Answer 4

Naive Bayers.

Question 5

What is the two-class classifier called which obtains the weight of the vector w, i.e, the linear decision boundary, that yields good class separation by using the Newton-Raphson algorithm to solve dE(w)/dw = 0.

Answer 5

Logistic Regression.

Question 5

What is the two-class classifier called which obtains the weight of the vector w, i.e, the linear decision boundary, that yields good class separation by using the Newton-Raphson algorithm to solve dE(w)/dw = 0.

Answer 5

Logistic Regression.