An Experimental Comparison of Hybrid Algorithms for Bayesian Network Structure Learning

@inproceedings{Gasse2012AnEC,
  title={An Experimental Comparison of Hybrid Algorithms for Bayesian Network Structure Learning},
  author={Maxime Gasse and Alex Aussem and Haytham Elghazel},
  booktitle={ECML/PKDD},
  year={2012}
}
We present a novel hybrid algorithm for Bayesian network structure learning, called Hybrid HPC (H2PC). It first reconstructs the skeleton of a Bayesian network and then performs a Bayesian-scoring greedy hill-climbing search to orient the edges. It is based on a subroutine called HPC, that combines ideas from incremental and divide-and-conquer constraint-based methods to learn the parents and children of a target variable. We conduct an experimental comparison of H2PC against Max-Min Hill… 
New data-driven approaches to improve probabilistic model structure learning
TLDR
A transfer learning approach to learn the structure of a Sum-Product Network which can be converted to a Bayesian network under certain conditions is proposed, and a weighted maximum satisfiability (MAX-SAT) problem is formulated to solve the V-structures selection problem.
Benchmarking dynamic Bayesian network structure learning algorithms
TLDR
A novel approach to generate standard DBNs based on tiling and novel technique of evaluation, adapted from the "static" Structural Hamming Distance proposed for Bayesian networks is proposed.
Learning big Gaussian Bayesian networks: partition, estimation, and fusion
TLDR
A divide-and-conquer framework, called partition-estimation-fusion (PEF), for structure learning of big networks with thousands or more of nodes but fewer samples, which can improve the accuracy of structure learning by 20% or more, while reducing running time up to two orders-of-magnitude.
Efficient score-based Markov Blanket discovery
Probabilistic Graphical Model Structure Learning: Application to Multi-Label Classification. (Apprentissage de Structure de Modèles Graphiques Probabilistes: Application à la Classification Multi-Label)
TLDR
This thesis proposes a new hybrid structure learning method, H2PC (Hybrid Hybrid Parents and Children), which combines a constraint-based approach (statistical independence tests) with a score- based approach (posterior probability of the structure), and establishes several theoretical results to characterize the ILFs based on the compositional graphoid axioms.
Modular network construction using eQTL data: an analysis of computational costs and benefits
TLDR
It is suggested that guided comprehensive searches for parsimonious networks should be considered as an alternative to greedy network searches.
Reasoning about Independence in Probabilistic Models of Relational Data
TLDR
This work provides a new representation, the abstract ground graph, that enables a sound, complete, and computationally ecient method for answering d-separation queries about relational models, and presents empirical results that demonstrate eectiveness.
...
...

References

SHOWING 1-10 OF 44 REFERENCES
An Efficient and Scalable Algorithm for Local Bayesian Network Structure Discovery
TLDR
An efficient and scalable constraint-based algorithm, called Hybrid Parents and Children (HPC), to learn the parents and children of a target variable in a Bayesian network, which combines ideas from incremental and divide-and-conquer methods in a principled and effective way.
The max-min hill-climbing Bayesian network structure learning algorithm
TLDR
The first empirical results simultaneously comparing most of the major Bayesian network algorithms against each other are presented, namely the PC, Sparse Candidate, Three Phase Dependency Analysis, Optimal Reinsertion, Greedy Equivalence Search, and Greedy Search.
Optimized Algorithm for Learning Bayesian Network Super-structures
TLDR
The OptHPC is presented, an optimized version of HPC that implements several optimizations to get an efficient global method for learning SS and shows the most favorable balance sensitivity/specificity and computational cost for use as super-structure estimator when compared to several state-of-the-art methods.
Finding Optimal Bayesian Network Given a Super-Structure
TLDR
A super-structure constrained optimal search (COS) is developed, which is an undirected graph that restricts the search to networks whose skeleton is a subgraph of S and can be approximated by IT approaches, enabling to control the trade-off between speed and accuracy.
A Simple Approach for Finding the Globally Optimal Bayesian Network Structure
TLDR
It is shown that it is possible to learn the best Bayesian network structure with over 30 variables, which covers many practically interesting cases and offers a possibility for efficient exploration of the best networks consistent with different variable orderings.
Optimal Search on Clustered Structural Constraint for Learning Bayesian Network Structure
TLDR
Empirical evaluations demonstrate that the proposed algorithm can learn optimal Bayesian networks for some graphs containing several hundreds of vertices, and even for super-structures having a high average degree, which is a drastic improvement in feasibility over the previous optimal algorithm.
Learning Bayesian Network Structure from Massive Datasets: The "Sparse Candidate" Algorithm
TLDR
An algorithm that achieves faster learning by restricting the search space, which restricts the parents of each variable to belong to a small subset of candidates and is evaluated both on synthetic and real-life data.
Exact Bayesian Structure Discovery in Bayesian Networks
TLDR
This work presents an algorithm that computes the exact posterior probability of a subnetwork, e.g., a directed edge, and shows that also in domains with a large number of variables, exact computation is feasible, given suitable a priori restrictions on the structures.
Optimal Reinsertion: A New Search Operator for Accelerated and More Accurate Bayesian Network Structure Learning
TLDR
A new algorithm called ORSearch is introduced which allows each optimal reinsertion step to be computed efficiently on large datasets and compares Optimal Reinsertion against a highly tuned implementation of multirestart hill climbing.
Optimal Structure Identification With Greedy Search
TLDR
This paper proves the so-called "Meek Conjecture", which shows that if a DAG H is an independence map of another DAG G, then there exists a finite sequence of edge additions and covered edge reversals in G such that H remains anindependence map of G and after all modifications G =H.
...
...