Description

The PAVER reports used in the paper: “Alternative Regularizations for Outer-Approximation Algorithms for Convex MINLP” by D.E. Bernal, Z. Peng, J. Kronqvist and I.E. Grossmann are available here.

Algorithms/Methods

The following algorithms/methods are benchmarked.

• OA
• ROA-$\ell_1$
• ROA-$\ell_2^2$
• ROA-$\ell_\infty$
• ROA-$\mathcal{L}_2$
• ROA-$\nabla_2 \mathcal{L}$
• ROA-$\mathcal{L}_1 /\ell_2^2$
• ROA-$\mathcal{L}_1$
• LP/NLP
• RLP/NLP-$\ell_1$
• RLP/NLP-$\ell_2^2$
• RLP/NLP-$\ell_\infty$
• RLP/NLP-$\mathcal{L}_2$
• RLP/NLP-$\nabla_2 \mathcal{L}$
• RLP/NLP-$\mathcal{L}_1 /\ell_2^2$
• RLP/NLP-$\mathcal{L}_1$

Instances

There are two problem set in this benchmark, Problem Set 1 and Problem Set 2.

• Problem Set 1 constains 438 convex instances from MINLPLib, which have at least one discrete variable and at least one continuous variable.
• Problem Set 2 contains 135 highly nonlinear instances.

Paver Reports

ROA (Regularized Outer Approximation) method and RLP/NLP(Regularized LP/NLP based Branch & Bound).

1. ROA benchmark in Problem Set 1

2. RLP/NLP benchmark in Problem Set 1

3. ROA benchmark in Problem Set 2

4. RLP/NLP benchmark in Problem Set 2

The reports can also be downloaded from the repository.