Sample Exam 1

Question 1

For SVM problem, identify support vectors

Answer 1

Vectors for which eigen values are non zero

Question 2

Calculate W for 2 class linearly separable case

Answer 2

W = sum( λ_i* y_i*x_i) for i eigenvalues/vectors (non zero eigenvalues)

Question 3

Calculating w₀ for svm

Answer 3

Use equation:

y_i(W^T*x_i + w₀ ) = 1

Question 4

What is meant by first 2 principal components for PCA

Answer 4

2 largest eigenvalues

Question 5

Describe what sparse coding does

Answer 5

Feature extraction
Projects data into a few non random directions in a high dimensional space