Open-Source Software & Benchmarks
SECQUOIA develops and maintains open-source solvers, algebraic reformulation packages, and curated benchmark suites across Python (Pyomo) and Julia (JuMP / JuliaQUBO). Each tool below includes repository links, documentation, licensing, and reference papers for scholarly citation.
MindtPy
The Mixed-Integer Nonlinear Decomposition Toolbox in Pyomo (MindtPy) is an open-source solver for Mixed-Integer Nonlinear Programming (MINLP) integrated directly into the Pyomo modeling ecosystem. MindtPy implements modular decomposition strategies that coordinate MILP and NLP subproblem solvers:
- Outer Approximation (OA) and Regularized Outer Approximation (ROA)
- LP/NLP-based Single-Tree Branch-and-Bound (LP/NLP BB)
- Global Outer Approximation (GOA) and Extended Cutting Plane (ECP)
- Feasibility Pump (FP) primal heuristics
Reference papers: Bernal et al., “Mixed-Integer Nonlinear Decomposition Toolbox for Pyomo (MindtPy),” Computer Aided Chemical Engineering 44, 895–900 (2018), doi:10.1016/B978-0-444-64241-7.50144-0; Bernal et al., Journal of Global Optimization 84, 807–842 (2022), doi:10.1007/s10898-022-01178-4.
QUBO.jl & ToQUBO.jl
QUBO.jl and ToQUBO.jl provide a unified Julia ecosystem for automatically reformulating, analyzing, benchmarking, and solving mathematical optimization models as Quadratic Unconstrained Binary Optimization (QUBO) and Ising instances via MathOptInterface.jl and JuMP.jl:
- Automatic compiler from constrained integer, continuous, and disjunctive JuMP models to QUBO (
ToQUBO.jl) with tunable variable encodings and penalty tightening - Standardized file I/O, instance analysis, and visualization tools (
QUBOTools.jl) - Unified solver wrappers for quantum annealers (D-Wave), coherent/classical heuristics, Simulated Bifurcation, and exact baselines (
QUBODrivers.jl)
Reference papers: Maciel Xavier et al., “QUBO.jl: A Julia Ecosystem for Quadratic Unconstrained Binary Optimization,” Optimization Methods and Software (2026), doi:10.1080/10556788.2026.2702926; Maciel Xavier et al., “Disjunctive Programming meets QUBO,” Computer Aided Chemical Engineering 53, 3433–3438 (2024), doi:10.1016/B978-0-443-28824-1.50573-1.
GDPLib
GDPLib is an open-source benchmark collection of Generalized Disjunctive Programming (GDP) models implemented in Pyomo. Covering process superstructure synthesis (such as HDA, catalytic distillation, Kaibel columns, membrane cascades, and heat-exchanger networks), scheduling, and logic-constrained engineering design, GDPLib supports reproducible benchmarking of logic-based solvers and reformulations.
Reference works: Ovalle et al., “Logic-Based Discrete-Steepest Descent: A solution method for process synthesis Generalized Disjunctive Programs,” Computers & Chemical Engineering 195, 108993 (2025), doi:10.1016/j.compchemeng.2024.108993; Lee, Kim & Bernal Neira, arXiv:2611.05508 (2026).
Pyomo.GDP & GDPOpt
Pyomo.GDP extends Pyomo with native algebraic representations of logical disjunctions and Boolean propositions, paired with automatic Big-M, Hull, and cutting-plane transformations as well as the GDPOpt logic-based outer-approximation (LOA), global logic-based (GLOA), and discrete-steepest-descent (LDSA) solver suite.
Reference paper: Chen et al., “Pyomo.GDP: an ecosystem for logic based modeling and optimization development,” Optimization and Engineering 23, 607–642 (2022), doi:10.1007/s11081-021-09601-7.
Benchmark Libraries & Additional Solvers
- QOBLIB — Quantum Optimization Benchmarking Library (CC-BY-4.0, Co-Author): Curated, hard combinatorial optimization benchmark suite with standardized instance formats and verified classical reference bounds. Cite Koch et al., Nature Computational Science (2026), doi:10.1038/s43588-026-00991-1.
- HamLib — Library of Hamiltonians (CC-BY-4.0, Co-Author): Large-scale dataset of condensed-matter, chemistry, and discrete-optimization Hamiltonians for benchmarking quantum simulation and optimization hardware. Cite Sawaya et al., Quantum 8, 1559 (2024), doi:10.22331/q-2024-12-11-1559.
- PySA & Stochastic-Benchmark (Open Source, Contributor / Developer): High-performance simulated annealing / parallel tempering in Python (NASA) and statistical solver-evaluation framework (USRA-RIACS / NASA QuAIL).
- DICOPT (GAMS Commercial MINLP Solver, Contributor): Feasibility-pump enhancements for convex and nonconvex MINLP; cite Bernal et al., Optimization Methods and Software 35(1), 171–190 (2020), doi:10.1080/10556788.2019.1641498.