-
Learning Dynamic Bayesian Networks from Data: Foundations, First Principles and Numerical Comparisons
Authors:
Vyacheslav Kungurtsev,
Petr Rysavy,
Fadwa Idlahcen,
Pavel Rytir,
Ales Wodecki
Abstract:
In this paper, we present a guide to the foundations of learning Dynamic Bayesian Networks (DBNs) from data in the form of multiple samples of trajectories for some length of time. We present the formalism for a generic as well as a set of common types of DBNs for particular variable distributions. We present the analytical form of the models, with a comprehensive discussion on the interdependence…
▽ More
In this paper, we present a guide to the foundations of learning Dynamic Bayesian Networks (DBNs) from data in the form of multiple samples of trajectories for some length of time. We present the formalism for a generic as well as a set of common types of DBNs for particular variable distributions. We present the analytical form of the models, with a comprehensive discussion on the interdependence between structure and weights in a DBN model and their implications for learning. Next, we give a broad overview of learning methods and describe and categorize them based on the most important statistical features, and how they treat the interplay between learning structure and weights. We give the analytical form of the likelihood and Bayesian score functions, emphasizing the distinction from the static case. We discuss functions used in optimization to enforce structural requirements. We briefly discuss more complex extensions and representations. Finally we present a set of comparisons in different settings for various distinct but representative algorithms across the variants.
△ Less
Submitted 25 June, 2024;
originally announced June 2024.
-
ExDAG: Exact learning of DAGs
Authors:
Pavel Rytíř,
Aleš Wodecki,
Jakub Mareček
Abstract:
There has been a growing interest in causal learning in recent years. Commonly used representations of causal structures, including Bayesian networks and structural equation models (SEM), take the form of directed acyclic graphs (DAGs). We provide a novel mixed-integer quadratic programming formulation and associated algorithm that identifies DAGs on up to 50 vertices, where these are identifiable…
▽ More
There has been a growing interest in causal learning in recent years. Commonly used representations of causal structures, including Bayesian networks and structural equation models (SEM), take the form of directed acyclic graphs (DAGs). We provide a novel mixed-integer quadratic programming formulation and associated algorithm that identifies DAGs on up to 50 vertices, where these are identifiable. We call this method ExDAG, which stands for Exact learning of DAGs. Although there is a superexponential number of constraints that prevent the formation of cycles, the algorithm adds constraints violated by solutions found, rather than imposing all constraints in each continuous-valued relaxation. Our empirical results show that ExDAG outperforms local state-of-the-art solvers in terms of precision and outperforms state-of-the-art global solvers with respect to scaling, when considering Gaussian noise. We also provide validation with respect to other noise distributions.
△ Less
Submitted 21 June, 2024;
originally announced June 2024.
-
Scheduling a Multi-Product Pipeline: A Discretized MILP Formulation
Authors:
Ales Wodecki,
Pavel Rytir,
Vyacheslav Kungurtsev,
Jakub Marecek
Abstract:
Multi-product pipelines are a highly efficient means of transporting liquids. Traditionally used to transport petroleum, its products and derivatives, they are now being repurposed to transport liquified natural gas admixed with hydrogen of various colors. We propose a novel mixed-integer linear programming (MILP) formulation, which optimizes efficiency while satisfying a wide range of real-world…
▽ More
Multi-product pipelines are a highly efficient means of transporting liquids. Traditionally used to transport petroleum, its products and derivatives, they are now being repurposed to transport liquified natural gas admixed with hydrogen of various colors. We propose a novel mixed-integer linear programming (MILP) formulation, which optimizes efficiency while satisfying a wide range of real-world constraints developed to meet the needs of the Czech national pipeline operator CEPRO. We provide tests on well-known synthetic (path-graph) networks and demonstrate the formulation's scaling properties using open-source and commercial MILP solvers.
△ Less
Submitted 18 December, 2023;
originally announced December 2023.
-
Binary Linear Codes, Dimers and Hypermatrices
Authors:
Martin Loebl,
Pavel Rytíř
Abstract:
We show that the weight enumerator of any binary linear code is equal to the permanent of a 3-dimensional hypermatrix (3-matrix). We also show that each permanent is a determinant of a 3-matrix. As an application we write the dimer partition function of a finite 3-dimensional cubic lattice as the determinant of the vertex-adjacency 3-matrix of a 2-dimensional simplicial complex which preserves the…
▽ More
We show that the weight enumerator of any binary linear code is equal to the permanent of a 3-dimensional hypermatrix (3-matrix). We also show that each permanent is a determinant of a 3-matrix. As an application we write the dimer partition function of a finite 3-dimensional cubic lattice as the determinant of the vertex-adjacency 3-matrix of a 2-dimensional simplicial complex which preserves the natural embedding of the cubic lattice.
△ Less
Submitted 26 September, 2016; v1 submitted 7 February, 2013;
originally announced February 2013.
-
Geometric representations of binary codes embeddable in three dimensions
Authors:
Pavel Rytíř
Abstract:
We say that a binary linear code C has a geometric representation if there exists a two dimensional simplicial complex D such that C is a punctured code of the kernel ker D of the incidence matrix of D and dim C = dim ker D. We show that every binary linear code has a geometric representation that can be embedded into R^4. Moreover, we show that a binary linear code C has a geometric representatio…
▽ More
We say that a binary linear code C has a geometric representation if there exists a two dimensional simplicial complex D such that C is a punctured code of the kernel ker D of the incidence matrix of D and dim C = dim ker D. We show that every binary linear code has a geometric representation that can be embedded into R^4. Moreover, we show that a binary linear code C has a geometric representation in R^3 if and only if there exists a graph G such that C equals the cut space of G. This is a polynomially testable property and hence we can conclude that there is a polynomial algorithm that decides the minimal dimension of a geometric representation of a binary linear code.
△ Less
Submitted 5 December, 2012;
originally announced December 2012.
-
Geometric representations of linear codes
Authors:
Pavel Rytíř
Abstract:
We say that a linear code C over a field F is triangular representable if there exists a two dimensional simplicial complex $Δ$ such that C is a punctured code of the kernel ker $Δ$ of the incidence matrix of $Δ$ over F and there is a linear map** between C and ker $Δ$ which is a bijection and maps minimal codewords to minimal codewords. We show that the linear codes over rationals and over GF(p…
▽ More
We say that a linear code C over a field F is triangular representable if there exists a two dimensional simplicial complex $Δ$ such that C is a punctured code of the kernel ker $Δ$ of the incidence matrix of $Δ$ over F and there is a linear map** between C and ker $Δ$ which is a bijection and maps minimal codewords to minimal codewords. We show that the linear codes over rationals and over GF(p), where p is a prime, are triangular representable. In the case of finite fields, we show that this representation determines the weight enumerator of C. We present one application of this result to the partition function of the Potts model.
On the other hand, we show that there exist linear codes over any field different from rationals and GF(p), p prime, that are not triangular representable. We show that every construction of triangular representation fails on a very weak condition that a linear code and its triangular representation have to have the same dimension.
△ Less
Submitted 5 May, 2011; v1 submitted 17 August, 2010;
originally announced August 2010.
-
Geometric representation of binary codes and computation of weight enumerators
Authors:
Pavel Rytíř
Abstract:
For every linear binary code $C$, we construct a geometric triangular configuration $Δ$ so that the weight enumerator of $C$ is obtained by a simple formula from the weight enumerator of the cycle space of $Δ$. The triangular configuration $Δ$ thus provides a geometric representation of $C$ which carries its weight enumerator. This is the initial step in the suggestion by M. Loebl, to extend the…
▽ More
For every linear binary code $C$, we construct a geometric triangular configuration $Δ$ so that the weight enumerator of $C$ is obtained by a simple formula from the weight enumerator of the cycle space of $Δ$. The triangular configuration $Δ$ thus provides a geometric representation of $C$ which carries its weight enumerator. This is the initial step in the suggestion by M. Loebl, to extend the theory of Pfaffian orientations from graphs to general linear binary codes. Then we carry out also the second step by constructing, for every triangular configuration $Δ$, a triangular configuration $Δ'$ and a bijection between the cycle space of $Δ$ and the set of the perfect matchings of $Δ'$.
△ Less
Submitted 10 July, 2009; v1 submitted 12 May, 2008;
originally announced May 2008.
