Saved in:
Bibliographic Details
Main Authors: Vlastelica, Marin, Paulus, Anselm, Musil, Vít, Martius, Georg, Rolínek, Michal
Format: Preprint
Published: 2019
Subjects:
Online Access:https://arxiv.org/abs/1912.02175
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910743559929856
author Vlastelica, Marin
Paulus, Anselm
Musil, Vít
Martius, Georg
Rolínek, Michal
author_facet Vlastelica, Marin
Paulus, Anselm
Musil, Vít
Martius, Georg
Rolínek, Michal
contents Achieving fusion of deep learning with combinatorial algorithms promises transformative changes to artificial intelligence. One possible approach is to introduce combinatorial building blocks into neural networks. Such end-to-end architectures have the potential to tackle combinatorial problems on raw input data such as ensuring global consistency in multi-object tracking or route planning on maps in robotics. In this work, we present a method that implements an efficient backward pass through blackbox implementations of combinatorial solvers with linear objective functions. We provide both theoretical and experimental backing. In particular, we incorporate the Gurobi MIP solver, Blossom V algorithm, and Dijkstra's algorithm into architectures that extract suitable features from raw inputs for the traveling salesman problem, the min-cost perfect matching problem and the shortest path problem. The code is available at https://github.com/martius-lab/blackbox-backprop.
format Preprint
id arxiv_https___arxiv_org_abs_1912_02175
institution arXiv
publishDate 2019
record_format arxiv
spellingShingle Differentiation of Blackbox Combinatorial Solvers
Vlastelica, Marin
Paulus, Anselm
Musil, Vít
Martius, Georg
Rolínek, Michal
Machine Learning
Achieving fusion of deep learning with combinatorial algorithms promises transformative changes to artificial intelligence. One possible approach is to introduce combinatorial building blocks into neural networks. Such end-to-end architectures have the potential to tackle combinatorial problems on raw input data such as ensuring global consistency in multi-object tracking or route planning on maps in robotics. In this work, we present a method that implements an efficient backward pass through blackbox implementations of combinatorial solvers with linear objective functions. We provide both theoretical and experimental backing. In particular, we incorporate the Gurobi MIP solver, Blossom V algorithm, and Dijkstra's algorithm into architectures that extract suitable features from raw inputs for the traveling salesman problem, the min-cost perfect matching problem and the shortest path problem. The code is available at https://github.com/martius-lab/blackbox-backprop.
title Differentiation of Blackbox Combinatorial Solvers
topic Machine Learning
url https://arxiv.org/abs/1912.02175