[Coin-announce] Minotaur in COIN-OR

Ashutosh Mahajan amahajan at iitb.ac.in
Mon Jan 2 03:20:28 EST 2023

We are pleased to announce that Minotaur (https://github.com/coin-or/minotaur)
is now part of the COIN-OR family. Special thanks to Stefan Vigerske for
helping us with this move.

Minotaur is a toolkit for Mixed-Integer Nonlinear Optimization developed
mainly in C++. The design is modular, so different components of algorithms
can be customized or replaced. Shared-memory parallel solvers for convex
MINLPs are available. Minotaur implements its own computational graph, can
automatically detect special structures such as those amenable to perspective
reformulations, and exploits features like function separability and lazy cuts
callbacks in a branch-and-cut framework.

A solver for Mixed Intger Quadratically Constrained Quadratic Problems
(MIQCQPs) is also available. It creates McCormick based relaxations in a
spatial branch-and-bound algorithm. Bound tightening techniques and heuristics
are included.

Minotaur's solvers can be invoked through interfaces like AMPL and Pyomo.
Contributions, issues, bug-reports, or general comments are most welcome.

On behalf of Team-Minotaur
Ashutosh Mahajan
IEOR, IIT Bombay

More information about the Coin-announce mailing list