Arithmetic Circuits and Neural Networks for Regular Matroids
By: Christoph Hertrich, Stefan Kober, Georg Loho
Potential Business Impact:
Finds best ways to pick items for matroids.
We prove that there exist uniform $(+,\times,/)$-circuits of size $O(n^3)$ to compute the basis generating polynomial of regular matroids on $n$ elements. By tropicalization, this implies that there exist uniform $(\max,+,-)$-circuits and ReLU neural networks of the same size for weighted basis maximization of regular matroids. As a consequence in linear programming theory, we obtain a first example where taking the difference of two extended formulations can be more efficient than the best known individual extended formulation of size $O(n^6)$ by Aprile and Fiorini. Such differences have recently been introduced as virtual extended formulations. The proof of our main result relies on a fine-tuned version of Seymour's decomposition of regular matroids which allows us to identify and maintain graphic substructures to which we can apply a local version of the star-mesh transformation.
Similar Papers
Efficient Algorithms for Minimal Matroid Extensions and Irreducible Decompositions of Circuit Varieties
Combinatorics
Finds smallest building blocks of complex math structures.
Optimal Parallel Basis Finding in Graphic and Related Matroids
Data Structures and Algorithms
Finds important paths in networks faster.
A note on embracing exchange sequences in oriented matroids
Combinatorics
Finds shortest path between geometric shapes.