04.02.2015 Views

Stochastic Programming - Index of

Stochastic Programming - Index of

Stochastic Programming - Index of

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

RECOURSE PROBLEMS 165<br />

for feasibility. We should like to check for feasibility in such a way that if the<br />

given problem is not feasible, we automatically come up with a generator <strong>of</strong><br />

pol pos W . For the discussion, we shall use Figure 5.<br />

We should like to find a σ such that<br />

σ T t ≤ 0 for all t ∈ pos W.<br />

This is equivalent to requiring that σ T W ≤ 0. In other words, σ should be<br />

in the cone pol pos W . But, assuming that the right-hand side h(ξ) − T (ξ)ˆx<br />

produces an infeasible problem, we should at the same time require that<br />

σ T [h(ξ) − T (ξ)ˆx] > 0,<br />

because if we later add the constraint σ T [h(ξ) − T (ξ)x] ≤ 0 to our problem,<br />

we shall exclude the infeasible right-hand side h(ξ) − T (ξ)ˆx without leaving<br />

out any feasible solutions. Hence we should like to solve<br />

max<br />

σ<br />

{σT (h(ξ) − T (ξ)ˆx) | σ T W ≤ 0, ‖σ‖ ≤1},<br />

where the last constraint has been added to bound σ. We can do that, because<br />

otherwise the maximal value will be +∞, and that does not interest us since<br />

we are looking for the direction defined by σ. If we had chosen the l 2 norm, the<br />

maximization would have made sure that σ came as close to h(ξ) − T (ξ)ˆx as<br />

possible (see Figure 5). Computationally, however, we should not like to work<br />

with quadratic constraints. Let us therefore see what happens if we choose<br />

the l 1 norm. Let us write our problem differently to see the details better. To<br />

do that, we need to let the unconstrained σ be replaced by σ 1 − σ 2 ,where<br />

σ 1 ,σ 2 ≥ 0. We then get the following:<br />

max{(σ 1 −σ 2 ) T (h(ξ)−T (ξ)ˆx) | (σ 1 −σ 2 ) T W ≤ 0, e T (σ 1 +σ 2 ) ≤ 1, σ 1 ,σ 2 ≥ 0},<br />

where e is a vector <strong>of</strong> ones. To more easily find the dual <strong>of</strong> this problem, let<br />

us write it down in a more standard format:<br />

max(σ 1 − σ 2 ) T (h(ξ) − T (ξ)ˆx) dual variables<br />

W T σ 1 − W T σ 2 ≤ 0<br />

y<br />

e T σ 1 + e T σ 2 ≤ 1 t<br />

σ 1 ,σ 2 ≥ 0<br />

From this, we find the dual linear program to be<br />

min{t | Wy+ et ≥ (h(ξ) − T (ξ)ˆx), −Wy+ et ≥−(h(ξ) − T (ξ)ˆx), y,t ≥ 0}.<br />

Note that if the optimal value in this problem is zero, we have Wy =<br />

h(ξ) − T (ξ)ˆx, so that we do indeed have h(ξ) − T (ξ)ˆx ∈ pos W , contrary

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!