NO - Chapter 3 (Modified). How to solve a Linear Program

NO - Reduced Costs and Descent Direction
Summary:

The reduced cost is given by:
$(c_{N}^{t} - c_{B}^{t} B^{- 1} N) x_{N}$
Where:

$x_{N}$ is a vector and it’s equal to $0$ only if we have that $x$ is a vertex

We define the descent direction as a direction given by $x_{N}$ such that if we increase it the reduced cost decreases.

If we cannot find a new descent direction, then we are in a local optima, also if we have that the problem is convex we know that it’s the global optima

Link to original