LP and QP in conic form

The LP

 displaystylemin_x : f^Tx ~:~ Fx le g

can always be represented in the conic form

 displaystylemin : c^Tz ~:~ Az = b, ;; z ge 0,

for appropriate matrix A and vector c.

To prove this, we introduce three new sets of non-negative variables: s = g-Fx, which represents the constraints, and x_+,x_-, which contain the positive and negative parts of vector x. We have x = x_+-x_-, with x_+ := max(x,0) ge 0 and x_- = max(-x,0) ge 0. The constraint Fx le g then reads s ge 0. Let us define the new variable z = (x_+,x_-,s)/ The relationship between s,x_+,x_- can then be expressed as Az  = b, with

 A := left(begin{array}{ccc}-F & F & -I end{array}right), ;; b := g.

The objective of the original problem can be written as f^Tx = c^Tz, with

 c := left(begin{array}{c} f  -f  0 end{array}right).

Putting this together, we express the original LP as claimed