gtsam 4.1.1
gtsam
|
Policy for ActivetSetSolver to solve Linear Programming.
Static Public Member Functions | |
static GaussianFactorGraph | buildCostFunction (const LP &lp, const VectorValues &xk) |
Create the factor ||x-xk - (-g)||^2 where xk is the current feasible solution on the constraint surface and g is the gradient of the linear cost, i.e. More... | |
Static Public Attributes | |
static constexpr double | maxAlpha = std::numeric_limits<double>::infinity() |
Maximum alpha for line search x'=xk + alpha*p, where p is the cost gradient For LP, maxAlpha = Infinity. | |
|
inlinestatic |
Create the factor ||x-xk - (-g)||^2 where xk is the current feasible solution on the constraint surface and g is the gradient of the linear cost, i.e.
-g is the direction we wish to follow to decrease the cost.
Essentially, we try to match the direction d = x-xk with -g as much as possible subject to the condition that x needs to be on the constraint surface, i.e., d is along the surface's subspace.
The least-square solution of this quadratic subject to a set of linear constraints is the projection of the gradient onto the constraints' subspace