LaplaceKorea / ToQUBO.jl

🟦 JuMP ToQUBO Automatic Reformulation

ToQUBO.jl 🟥🟩🟪🟦

Introduction

ToQUBO.jl is a Julia package to reformulate general optimization problems into QUBO (Quadratic Unconstrained Binary Optimization) instances. This tool aims to convert a broad range of JuMP problems for straightforward application in many physics and physics-inspired solution methods whose normal optimization form is equivalent to the QUBO. These methods include quantum annealing, quantum gate-circuit optimization algorithms (Quantum Optimization Alternating Ansatz, Variational Quantum Eigensolver), other hardware-accelerated platforms, such as Coherent Ising Machines and Simulated Bifurcation Machines, and more traditional methods such as simulated annealing. During execution, ToQUBO.jl encodes both discrete and continuous variables, maps constraints, and computes their penalties, performing a few model optimization steps along the process. A simple interface to connect various annealers and samplers as QUBO solvers is defined in Anneal.jl.

ToQUBO.jl was written as a MathOptInterface (MOI) layer that automatically maps between input and output models, thus providing a smooth JuMP modeling experience.

Getting Started

Installation

ToQUBO is available via Julia's Pkg:

julia> ]add ToQUBO

or

julia> using Pkg

julia> Pkg.add("ToQUBO")

Simple Example

using JuMP
using ToQUBO
using Anneal

model = Model(() -> ToQUBO.Optimizer(ExactSampler.Optimizer))

@variable(model, x[1:3], Bin)
@constraint(model, 0.3*x[1] + 0.5*x[2] + 1.0*x[3] <= 1.6)
@objective(model, Max, 1.0*x[1] + 2.0*x[2] + 3.0*x[3])

optimize!(model)

for i = 1:result_count(model)
    xᵢ = value.(x, result = i)
    yᵢ = objective_value(model, result = i)
    println("f($xᵢ) = $yᵢ")
end

List of Interpretable Constraints

Below, we present a list containing all⁴ MOI constraint types and their current reformulation support by ToQUBO.

Linear constraints

Mathematical Constraint	MOI Function	MOI Set	Status
$\vec{a}' \vec{x} \le \beta$	ScalarAffineFunction	LessThan	✔️
$\vec{a}' \vec{x} \ge \alpha$	ScalarAffineFunction	GreaterThan	♻️
$\vec{a}' \vec{x} = \beta$	ScalarAffineFunction	EqualTo	✔️
$\alpha \le \vec{a}' \vec{x} \le \beta$	ScalarAffineFunction	Interval	♻️
$x_i \le \beta$	VariableIndex	LessThan	✔️
$x_i \ge \alpha$	VariableIndex	GreaterThan	✔️
$x_i = \beta$	VariableIndex	EqualTo	✔️
$\alpha \le x_i \le \beta$	VariableIndex	Interval	✔️
$A \vec{x} + b \in \mathbb{R}_{+}^{n}$	VectorAffineFunction	Nonnegatives	♻️
$A \vec{x} + b \in \mathbb{R}_{-}^{n}$	VectorAffineFunction	Nonpositives	♻️
$A \vec{x} + b = 0$	VectorAffineFunction	Zeros	♻️

Conic constraints

Mathematical Constraint	MOI Function	MOI Set	Status
$\left\lVert{}{A \vec{x} + b}\right\rVert{}_{2} \le \vec{c}' \vec{x} + d$	VectorAffineFunction	SecondOrderCone	📖
$y \ge \left\lVert{}{\vec{x}}\right\rVert{}_{2}$	VectorOfVariables	SecondOrderCone	📖
$2 y z \ge \left\lVert{}{\vec{x}}\right\rVert{}_{2}^{2}; y, z \ge 0$	VectorOfVariables	RotatedSecondOrderCone	📖
$\left( \vec{a}'_1 \vec{x} + b_1,\vec{a}'_2 \vec{x} + b_2,\vec{a}'_3 \vec{x} + b_3 \right) \in E$	VectorAffineFunction	ExponentialCone	❌
$A(\vec{x}) \in S_{+}$	VectorAffineFunction	PositiveSemidefiniteConeTriangle	❌
$B(\vec{x}) \in S_{+}$	VectorAffineFunction	PositiveSemidefiniteConeSquare	❌
$\vec{x} \in S_{+}$	VectorOfVariables	PositiveSemidefiniteConeTriangle	❌
$\vec{x} \in S_{+}$	VectorOfVariables	PositiveSemidefiniteConeSquare	❌

Quadratic constraints

Mathematical Constraint	MOI Function	MOI Set	Status
$\vec{x} Q \vec{x} + \vec{a}' \vec{x} + b \ge 0$	ScalarQuadraticFunction	GreaterThan	♻️
$\vec{x} Q \vec{x} + \vec{a}' \vec{x} + b \le 0$	ScalarQuadraticFunction	LessThan	✔️
$\vec{x} Q \vec{x} + \vec{a}' \vec{x} + b = 0$	ScalarQuadraticFunction	EqualTo	✔️
Bilinear matrix inequality	VectorQuadraticFunction	PositiveSemidefiniteCone	❌

Discrete and logical constraints

Mathematical Constraint	MOI Function	MOI Set	Status
$x_i \in \mathbb{Z}$	VariableIndex	Integer	✔️
$x_i \in \left\lbrace{0, 1}\right\rbrace$	VariableIndex	ZeroOne	✔️
$x_i \in \left\lbrace{0}\right\rbrace \cup \left[{l, u}\right]$	VariableIndex	Semicontinuous	⌛
$x_i \in \left\lbrace{0}\right\rbrace \cup \left[{l, l + 1, \dots, u - 1, u}\right]$	VariableIndex	Semiinteger	⌛
¹	VectorOfVariables	SOS1	✔️
²	VectorOfVariables	SOS2	📖
$y = 1 \implies \vec{a}' \vec{x} \in S$	VectorAffineFunction	Indicator	📖

¹ At most one component of x can be nonzero

² At most two components of x can be nonzero, and if so they must be adjacent components

Symbol	Meaning
✔️	Available
♻️	Available through Bridges³
❌	Unavailable
⌛	Under Development (Available soon)
📖	Under Research

³ MOI Bridges provide equivalent constraint mapping.

⁴ If you think this list is incomplete, consider creating an Issue or opening a Pull Request.

Citing ToQUBO.jl

If you use ToQUBO.jl in your work, we kindly ask you to include the following citation:

@software{toqubo:2022,
  author       = {Pedro Xavier and Tiago Andrade and Joaquim Garcia and David Bernal},
  title        = {{ToQUBO.jl}},
  month        = {mar},
  year         = {2022},
  publisher    = {Zenodo},
  version      = {v0.1.0},
  doi          = {10.5281/zenodo.6387592},
  url          = {https://doi.org/10.5281/zenodo.6387592}
}

PSR Quantum Optimization Toolchain