An efficient modeling interface in Python with support for HIGHS

[metab0t]

Today, I want to share a newly developed modeling interface for Python: https://github.com/metab0t/PyOptInterface with support for HiGHS.

It can be installed with our PyPI package of HiGHS to use HiGHS in Python out of the box:

pip install pyoptinterface[highs]

It provides high-level modeling capability by using the C API of HiGHS directly and achieves competitive performance to construct large scale models.

It has built-in support for N-dimensional containers for variables and constraints. The variables and constraints can also be used with Numpy to formulate multi-dimensional array.

This is an introductory example to use PyOptInterface and HiGHS to solve N-queens problem:

import numpy as np
import pyoptinterface as poi
from pyoptinterface import highs

model = highs.Model()

N = 8

x = np.empty((N, N), dtype=object)
for i in range(N):
    for j in range(N):
        x[i, j] = model.add_variable(domain=poi.VariableDomain.Binary)

for i in range(N):
    # Row and column
    model.add_linear_constraint(poi.quicksum(x[i, :]), poi.Eq, 1.0)
    model.add_linear_constraint(poi.quicksum(x[:, i]), poi.Eq, 1.0)
for i in range(-N+1, N):
    # Diagonal
    model.add_linear_constraint(poi.quicksum(x.diagonal(i)), poi.Leq, 1.0)
    # Anti-diagonal
    model.add_linear_constraint(poi.quicksum(np.fliplr(x).diagonal(i)), poi.Leq, 1.0)

model.optimize()

print("Termination status:", model.get_model_attribute(poi.ModelAttribute.TerminationStatus))

get_v = np.vectorize(lambda x: model.get_value(x))
x_value = get_v(x)

print(x_value.astype(int))

More detailed documentation is available online.

Thanks for @jajhall for developing such an amazing open source optimizer!

Feedbacks and stars are all welcome 😄.

[jajhall]

Thanks for developing this Python packaging of HiGHS. I didn't find any documentation for the methods in the API.

Note that HiGHS would not have been developed without @galabovaa, or the writers of its solvers: Qi Huangfu, @lschork2, @lgottwald and @feldmeier.

[metab0t]

Yep, the autogenerated API documentation is still sparse.

The main APIs for modeling interface are documented on this page

The Python packaging of HiGHS enables users to have instant access to HiGHS without downloading binaries manually.

[jajhall]

OK, thanks. I understand why you're focusing on the top level modelling interface