Differentiating a QP wrt a single variable

Consider the quadratic program

\[\begin{split} \begin{array} {ll} \mbox{minimize} & \frac{1}{2} x^T Q x + q^T x \\ \mbox{subject to} & G x \leq h, x \in \mathcal{R}^2 \\ \end{array} \end{split}\]

where Q, q, G are fixed and h is the single parameter.

In this example, we'll try to differentiate the QP wrt h, by finding its jacobian by hand (using Eqn (6) of QPTH article) and compare the results:

  • Manual compuation
  • Using JuMP and DiffOpt

Assuming

Q = [[4, 1], [1, 2]]
q = [1, 1]
G = [1, 1]

and begining with a starting value of h=-1

few values just for reference

variableoptimal valuenote
x*[-0.25; -0.75]Primal optimal
𝜆∗-0.75Dual optimal

Finding Jacobian using matrix inversion

Lets formulate Eqn (6) of QPTH article for our QP. If we assume h as the only parameter and Q,q,G as fixed problem data - also note that our QP doesn't involves Ax=b constraint - then Eqn (6) reduces to

\[\begin{gather} \begin{bmatrix} Q & g^T \\ \lambda^* G & G x^* - h \end{bmatrix} \begin{bmatrix} dx \\ d \lambda \end{bmatrix} = \begin{bmatrix} 0 \\ \lambda^* dh \end{bmatrix} \end{gather}\]

Now to find the jacobians $ \frac{\partial x}{\partial h}, \frac{\partial \lambda}{\partial h}$ we substitute dh = I = [1] and plug in values of Q,q,G to get

\[\begin{gather} \begin{bmatrix} 4 & 1 & 1 \\ 1 & 2 & 1 \\ -0.75 & -0.75 & 0 \end{bmatrix} \begin{bmatrix} \frac{\partial x_1}{\partial h} \\ \frac{\partial x_2}{\partial h} \\ \frac{\partial \lambda}{\partial h} \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \\ -0.75 \end{bmatrix} \end{gather}\]

Upon solving using matrix inversion, the jacobian is

\[\frac{\partial x_1}{\partial h} = 0.25, \frac{\partial x_2}{\partial h} = 0.75, \frac{\partial \lambda}{\partial h} = -1.75\]

Finding Jacobian using JuMP and DiffOpt

using JuMP
import DiffOpt
import Ipopt

n = 2 # variable dimension
m = 1; # no of inequality constraints

Q = [4.0 1.0; 1.0 2.0]
q = [1.0; 1.0]
G = [1.0 1.0;]
h = [-1.0;]   # initial values set
1-element Vector{Float64}:
 -1.0

Initialize empty model

model = Model(() -> DiffOpt.diff_optimizer(Ipopt.Optimizer))
set_silent(model)

Add the variables

@variable(model, x[1:2])
2-element Vector{JuMP.VariableRef}:
 x[1]
 x[2]

Add the constraints.

@constraint(model, cons[j in 1:1], sum(G[j, i] * x[i] for i in 1:2) <= h[j]);

@objective(
    model,
    Min,
    1 / 2 * sum(Q[j, i] * x[i] * x[j] for i in 1:2, j in 1:2) +
    sum(q[i] * x[i] for i in 1:2)
)

\[ 2 x_{1}^2 + x_{1}\times x_{2} + x_{2}^2 + x_{1} + x_{2} \]

Solve problem

optimize!(model)

primal solution

value.(x)
2-element Vector{Float64}:
 -0.24999999834694256
 -0.7499999950408276

dual solution

dual.(cons)
1-element Vector{Float64}:
 -0.7499999884285978

set sentivitity

MOI.set(
    model,
    DiffOpt.ForwardConstraintFunction(),
    cons[1],
    0.0 * index(x[1]) - 1.0,  # to indicate the direction vector to get directional derivatives
)

Note that 0.0 * index(x[1]) is used to make its type typeof(0.0 * index(x[1]) - 1.0) <: MOI.AbstractScalarFunction. To indicate different direction to get directional derivative, users should replace 0.0 * index(x[1]) - 1.0 as the form of dG*x - dh, where dG and dh correspond to the elements of direction vectors along G and h axes, respectively.

Compute derivatives

DiffOpt.forward_differentiate!(model)

Query derivative

dx = MOI.get.(model, DiffOpt.ForwardVariablePrimal(), x)
2-element Vector{Float64}:
 0.25000000385713417
 0.7500000115714026

This page was generated using Literate.jl.