Model quadratic objectives, inspect Hessians, and compare updates. Handle box limits with projected search automatically. Turn optimization inputs into practical decisions, tables, and exports.
This single-page tool solves two-variable quadratic convex programs with optional box constraints.
| q11 | q12 | q22 | c1 | c2 | Bounds | Start | Approx x* | Approx minimum |
|---|---|---|---|---|---|---|---|---|
| 4 | 1 | 3 | -8 | -6 | x1, x2 in [0, 5] | (0, 0) | (1.6364, 1.4545) | -10.9091 |
Objective: f(x1, x2) = 0.5(q11x12 + 2q12x1x2 + q22x22) + c1x1 + c2x2 + k
Gradient: ∇f(x) = [q11x1 + q12x2 + c1, q12x1 + q22x2 + c2]
Hessian: H = [[q11, q12], [q12, q22]]
Convexity rule: the model is convex when the Hessian is positive semidefinite. For this 2×2 symmetric case, the smallest eigenvalue must be nonnegative.
Projected update: x(k+1) = Π[l,u](x(k) − α∇f(x(k)))
Stopping check: the solver stops when both the step norm and projected gradient mapping norm drop below tolerance.
It solves two-variable quadratic optimization models with optional lower and upper bounds. The objective is evaluated with gradient-based iterative updates and convexity checks.
It computes the Hessian eigenvalues. When the smallest eigenvalue is nonnegative, the quadratic model is convex. A positive smallest eigenvalue means strong convexity.
Bounds keep the solution inside allowed ranges. This is useful for resource limits, nonnegative variables, and practical decision constraints.
It measures how close the current point is to satisfying first-order optimality under bounds. Smaller values indicate a better constrained solution.
Use it when you want faster progress on smooth convex models. If updates oscillate, reduce momentum or learning rate.
A practical starting point is about 1 divided by the largest Hessian eigenvalue. Smaller values improve stability but may slow convergence.
The stationary reference ignores bounds. If that point lies outside the allowed box, the constrained optimum moves to a feasible boundary or interior point.
This page is designed for two variables to keep inputs readable and outputs interpretable. The same concepts extend to larger numerical solvers.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.