Exercise 3.3 (Feasibility conditions)

Let x be an element of the standard form polyhedron P = {x ∈ ℜn∣Ax = b,x ≥0}. Prove that a vector d ∈ℜn is a feasible direction at x if and only if Ad = 0 and di ≥ 0 for every i such that xi = 0.

Answers

Proof. Let x ∈ P be arbitrary, i.e., Ax = b and x ≥0.

⟹

Suppose that d ∈ ℜn is feasible at x, i.e., by Definition 3.1, there exists a 𝜃 > 0 such that x + 𝜃d ∈ P. Then, we must have A(x + 𝜃d) = b. Since Ax = b, we obtain Ad = (b −Ax)∕𝜃 = 0. This covers the first assertion. Furthermore, x + 𝜃d is nonnegative, i.e., x + 𝜃d ≥0. At any index i such that xi = 0, we have xi + 𝜃di = 𝜃di ≥ 0. Dividing by the positive 𝜃 > 0, we get di ≥ 0, as desired.

⟸

Now suppose that we have a direction d ∈ ℜn such that Ad = 0 with di = 0 for every i such that xi = 0. We must find 𝜃 > 0 such that x + 𝜃d ≥ 0. We make the following observation:

i = 1,…,n : xi+𝜃di ≥ 0⟸ { di < 0,then 𝜃 ≤−xi di di ≥ 0,then no restriction

Inspired by this observation, choose an arbitrary 𝜃 with the property (cf. page 88, Eq. 3.2):

0 < 𝜃 < min ⁡ { −xi di| di < 0}.
(1)

(Notice that di≠0 implies xi≠0.) We now justify the above argument rigorously. For any xi, 1 ≤ i ≤ n, we have the following cases:

  • xi = 0, then di ≥ 0 by assumption, and we obtain xi + 𝜃di = 𝜃di ≥ 0.
  • xi > 0, then we have two more cases. If di ≥ 0, then xi + 𝜃di > 0. If, however, di < 0, then 𝜃 ≤−xi∕di and thus 𝜃di ≥−xi, as desired.

Thus, by choosing 𝜃 as in (1), we ensure that x + 𝜃d ≥ 0. Furthermore,

A (x + 𝜃d) = Ax + 𝜃Ad = b + 0 = b.

We conclude that d is a feasible direction at x.

□
User profile picture
2022-02-16 16:07
Comments