Cyber-Physical Systems

54  ◾  Cyber-Physical Systems

© 2010 Taylor & Francis Group, LLC

e support vector machine (SVM) scheme uses an optimization to find a maximum margin

hyperplane separating sets of points. To obtain equivalence, we associate μ

E

with a label 1. ere

is also a label –1 consisting of a feature expectation {μ(π

(j)

): j = 0,…, (i − 1). e desired vector is

ω

(i)

, a unit vector orthogonal to the maximum margin separating hyperplane. us, the SVM can

solve for ω

(i)

.

If t

(n+1)

≤ ∈ when the algorithm finishes, then the following result occurs due to the optimiza-

tion formula found in step 2:

∀ω with ‖ω‖

2

≤ 1 ∃

i

such that

ω

T

μ

(i)

≥ ω

T

μ

E

− ε.

One policy must be ...