Paper 3, Section II, H

Optimization
Part IB, 2015

Consider the linear programming problem PP :

 minimise cTx subject to Axb,x0,\text { minimise } c^{T} x \text { subject to } A x \geqslant b, x \geqslant 0,

where xx and cc are in Rn,A\mathbb{R}^{n}, A is a real m×nm \times n matrix, bb is in Rm\mathbb{R}^{m} and T{ }^{T} denotes transpose. Derive the dual linear programming problem DD. Show from first principles that the dual of DD is PP.

Suppose that cT=(6,10,11),bT=(1,1,3)c^{T}=(6,10,11), b^{T}=(1,1,3) and A=(138112244)A=\left(\begin{array}{lll}1 & 3 & 8 \\ 1 & 1 & 2 \\ 2 & 4 & 4\end{array}\right). Write down the dual DD and find the optimal solution of the dual using the simplex algorithm. Hence, or otherwise, find the optimal solution x=(x1,x2,x3)x^{*}=\left(x_{1}^{*}, x_{2}^{*}, x_{3}^{*}\right) of PP.

Suppose that cc is changed to c~=(6+ε1,10+ε2,11+ε3)\tilde{c}=\left(6+\varepsilon_{1}, 10+\varepsilon_{2}, 11+\varepsilon_{3}\right). Give necessary and sufficient conditions for xx^{*} still to be the optimal solution of PP. If ε1=ε2=0\varepsilon_{1}=\varepsilon_{2}=0, find the range of values for ε3\varepsilon_{3} for which xx^{*} is still the optimal solution of PP.