PiecewiseQuadratics.jl

PiecewiseQuadratics.jl is a Julia package for manipulation of univariate piecewise quadratic functions of the form

f(x) = p x² + q x + r, ∀ x ∈ [lb, ub]

where:

• p, q, r are scalar
• x is the decision variable
• lb is the lower bound of x
• ub is the upper bound of x

There is a particular focus on convex analysis and optimization. The following operations are supported:

• Addition
• Convex envelope
• Proximal operator
• Minimization
• Others, such as scalar multiplication, shifting, perspective transformation.

This package and SeparableOptimization.jl were originally developed by Nicholas Moehle, Ellis Brown, and Mykel Kochenderfer at BlackRock AI Labs. They were developed to produce the results in the following paper: arXiv:2103.05455.

• Installation
• Example
• Contribute

Use Julia's builtin package manager Pkg to install. From a Julia REPL:

] add PiecewiseQuadratics

We specify a bounded quadratic g(x) = x² + 2x + 4, ∀x ∈ [0,5] by

# (lb, ub, p, q, r)
g = BoundedQuadratic(0, 5, 1, 2, 4)

We specify a piecewise quadratic function by providing a list of bounded quadratics in order. Where the pieces overlap, we take the function value to be the minimum over all possible values.

We specify

as follows:

f = PiecewiseQuadratic([
  BoundedQuadratic(-Inf, 3.0, 1.0, -3.0, 3.0),
  BoundedQuadratic(3.0, 4.0, 0.0, -1.0, 3.0),
  BoundedQuadratic(4.0, 6.0, 2.0, -20.0, 47.0),
  BoundedQuadratic(6.0, 7.5, 0.0, 1.0, -7.0),
  BoundedQuadratic(7.5, Inf, 0.0, 4.0, -29.0)
])

Below is a plot of the above piece-wise quadratic and its convex envelope, and the code to generate it.

using Plots
plot(get_plot(f); grid=false, linestyle=:dash, color=:black, label="piece-wise quadratic")
plot!(get_plot(simplify(envelope(f))); color=:blue, la=0.5, label="envelope")

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate. See CONTRIBUTING.md for more detail.