Classification 2

Question 1

Multiclass Classification

Answer 1

• train multiple binary classifiers, one distinguishing a class from all others
• Different weights and/or features per class

Question 2

Block Feature Vectors

Answer 2

• feature vector for each label (health, sports, …).
• each feature vector contains “sections” for each label
• weight vector is derived by combining all feature vectors

Question 3

Different Feature Equation

Answer 3

argmax yEY w T f(x,y)

Question 4

Different Weight Equation

Answer 4

argmax yEY (w y) T f(x,y)

Question 5

Making Decisions

Answer 5

wT = combined feature vectors
w T f(x,y) = argmax on weight