-
Solving Mixed Integer Programs Using Neural Networks
Authors:
Vinod Nair,
Sergey Bartunov,
Felix Gimeno,
Ingrid von Glehn,
Pawel Lichocki,
Ivan Lobov,
Brendan O'Donoghue,
Nicolas Sonnerat,
Christian Tjandraatmadja,
Pengming Wang,
Ravichandra Addanki,
Tharindi Hapuarachchi,
Thomas Keck,
James Keeling,
Pushmeet Kohli,
Ira Ktena,
Yujia Li,
Oriol Vinyals,
Yori Zwols
Abstract:
Mixed Integer Programming (MIP) solvers rely on an array of sophisticated heuristics developed with decades of research to solve large-scale MIP instances encountered in practice. Machine learning offers to automatically construct better heuristics from data by exploiting shared structure among instances in the data. This paper applies learning to the two key sub-tasks of a MIP solver, generating…
▽ More
Mixed Integer Programming (MIP) solvers rely on an array of sophisticated heuristics developed with decades of research to solve large-scale MIP instances encountered in practice. Machine learning offers to automatically construct better heuristics from data by exploiting shared structure among instances in the data. This paper applies learning to the two key sub-tasks of a MIP solver, generating a high-quality joint variable assignment, and bounding the gap in objective value between that assignment and an optimal one. Our approach constructs two corresponding neural network-based components, Neural Diving and Neural Branching, to use in a base MIP solver such as SCIP. Neural Diving learns a deep neural network to generate multiple partial assignments for its integer variables, and the resulting smaller MIPs for un-assigned variables are solved with SCIP to construct high quality joint assignments. Neural Branching learns a deep neural network to make variable selection decisions in branch-and-bound to bound the objective value gap with a small tree. This is done by imitating a new variant of Full Strong Branching we propose that scales to large instances using GPUs. We evaluate our approach on six diverse real-world datasets, including two Google production datasets and MIPLIB, by training separate neural networks on each. Most instances in all the datasets combined have $10^3-10^6$ variables and constraints after presolve, which is significantly larger than previous learning approaches. Comparing solvers with respect to primal-dual gap averaged over a held-out set of instances, the learning-augmented SCIP is 2x to 10x better on all datasets except one on which it is $10^5$x better, at large time limits. To the best of our knowledge, ours is the first learning approach to demonstrate such large improvements over SCIP on both large-scale real-world application datasets and MIPLIB.
△ Less
Submitted 29 July, 2021; v1 submitted 23 December, 2020;
originally announced December 2020.
-
Patterns of conjunctive forks
Authors:
Vašek Chvátal,
František Matúš,
Yori Zwólš
Abstract:
Three events in a probability space form a conjunctive fork if they satisfy specific constraints on conditional independence and covariances. Patterns of conjunctive forks within collections of events are characterized by means of systems of linear equations that have positive solutions. This characterization allows patterns of conjunctive forks to be recognized in polynomial time. Relations to pr…
▽ More
Three events in a probability space form a conjunctive fork if they satisfy specific constraints on conditional independence and covariances. Patterns of conjunctive forks within collections of events are characterized by means of systems of linear equations that have positive solutions. This characterization allows patterns of conjunctive forks to be recognized in polynomial time. Relations to previous work on causal betweenness and on patterns of conditional independence among random variables are discussed.
△ Less
Submitted 29 August, 2016; v1 submitted 13 August, 2016;
originally announced August 2016.
-
Minimum length path decompositions
Authors:
Dariusz Dereniowski,
Wieslaw Kubiak,
Yori Zwols
Abstract:
We consider a bi-criteria generalization of the pathwidth problem, where, for given integers $k,l$ and a graph $G$, we ask whether there exists a path decomposition $\cP$ of $G$ such that the width of $\cP$ is at most $k$ and the number of bags in $\cP$, i.e., the \emph{length} of $\cP$, is at most $l$.
We provide a complete complexity classification of the problem in terms of $k$ and $l$ for ge…
▽ More
We consider a bi-criteria generalization of the pathwidth problem, where, for given integers $k,l$ and a graph $G$, we ask whether there exists a path decomposition $\cP$ of $G$ such that the width of $\cP$ is at most $k$ and the number of bags in $\cP$, i.e., the \emph{length} of $\cP$, is at most $l$.
We provide a complete complexity classification of the problem in terms of $k$ and $l$ for general graphs. Contrary to the original pathwidth problem, which is fixed-parameter tractable with respect to $k$, we prove that the generalized problem is NP-complete for any fixed $k\geq 4$, and is also NP-complete for any fixed $l\geq 2$. On the other hand, we give a polynomial-time algorithm that, for any (possibly disconnected) graph $G$ and integers $k\leq 3$ and $l>0$, constructs a path decomposition of width at most $k$ and length at most $l$, if any exists.
As a by-product, we obtain an almost complete classification of the problem in terms of $k$ and $l$ for connected graphs. Namely, the problem is NP-complete for any fixed $k\geq 5$ and it is polynomial-time for any $k\leq 3$. This leaves open the case $k=4$ for connected graphs.
△ Less
Submitted 12 February, 2013;
originally announced February 2013.
-
A De Bruijn-Erdos theorem for chordal graphs
Authors:
Laurent Beaudou,
Adrian Bondy,
Xiaomin Chen,
Ehsan Chiniforooshan,
Maria Chudnovsky,
Vasek Chvatal,
Nicolas Fraiman,
Yori Zwols
Abstract:
A special case of a combinatorial theorem of De Bruijn and Erdos asserts that every noncollinear set of n points in the plane determines at least n distinct lines. Chen and Chvatal suggested a possible generalization of this assertion in metric spaces with appropriately defined lines. We prove this generalization in all metric spaces induced by connected chordal graphs.
A special case of a combinatorial theorem of De Bruijn and Erdos asserts that every noncollinear set of n points in the plane determines at least n distinct lines. Chen and Chvatal suggested a possible generalization of this assertion in metric spaces with appropriately defined lines. We prove this generalization in all metric spaces induced by connected chordal graphs.
△ Less
Submitted 30 January, 2012;
originally announced January 2012.
-
Lines in hypergraphs
Authors:
Laurent Beaudou,
Adrian Bondy,
Xiaomin Chen,
Ehsan Chiniforooshan,
Maria Chudnovsky,
Vasek Chvatal,
Nicolas Fraiman,
Yori Zwols
Abstract:
One of the De Bruijn - Erdos theorems deals with finite hypergraphs where every two vertices belong to precisely one hyperedge. It asserts that, except in the perverse case where a single hyperedge equals the whole vertex set, the number of hyperedges is at least the number of vertices and the two numbers are equal if and only if the hypergraph belongs to one of simply described families, near-pen…
▽ More
One of the De Bruijn - Erdos theorems deals with finite hypergraphs where every two vertices belong to precisely one hyperedge. It asserts that, except in the perverse case where a single hyperedge equals the whole vertex set, the number of hyperedges is at least the number of vertices and the two numbers are equal if and only if the hypergraph belongs to one of simply described families, near-pencils and finite projective planes. Chen and Chvatal proposed to define the line uv in a 3-uniform hypergraph as the set of vertices that consists of u, v, and all w such that {u,v,w} is a hyperedge. With this definition, the De Bruijn - Erdos theorem is easily seen to be equivalent to the following statement: If no four vertices in a 3-uniform hypergraph carry two or three hyperedges, then, except in the perverse case where one of the lines equals the whole vertex set, the number of lines is at least the number of vertices and the two numbers are equal if and only if the hypergraph belongs to one of two simply described families. Our main result eneralizes this statement by allowing any four vertices to carry three hyperedges (but keeping two forbidden): the conclusion remains the same except that a third simply described family, complements of Steiner triple systems, appears in the extremal case.
△ Less
Submitted 1 December, 2011;
originally announced December 2011.
