Decision Boundary is 90 degree to vector theta in SVM
- sigmoid(theta’ * x) > 1/2, we classify as 1; otherwise as 0
- 1 / (1 + exp(-(theta’ * x))) = 1 / 2 is the decision boundary
- exp(-(theta’ * x)) = 1
- theta’ * x = 0
- theta_1 * x_1 + theta_2 * x_2 = 0
- So (theta_1, theta_2) is 90 degree to (x1, x2).
