Custom Penalty Solver

This is an implementation of solving constrained binary quadratic programs with custom penalties using the variational quantum eigensolver (VQE). The original work is available at arXiv:2604.20088.

Getting started

We recommend using uv (installation guide).

uv venv --python 3.12
uv pip install custom-penalty

Or you can go with the traditional:

python3.12 -m venv .venv
source .venv/bin/activate
pip install custom-penalty

Usage

It is easier to define the problem with docplex, then convert it into a QuadraticProgram:

mdl = Model()

# define variables
x = mdl.binary_var_list(2, name='x')

# define objectives
mdl.maximize(x[0] + x[1] - x[0]*x[1])

# add constraints
mdl.add_constraint(x[0] + x[1] <= 1)
mdl.add_constraint(2*x[0] + 3*x[1] == 2)
mdl.add_constraint(5*x[0] + 6*x[1] >= 5)

# convert to quadratic program
qp = from_docplex_mp(mdl)

Then, you can solve it using

cps = CustomPenaltySolver(qp)
result = cps.solve()
print(result)

For customization and more detailed guide, refer example.ipynb.