Learning statistical models from relational data

Getoor, Lise; Mihalkova, Lilyana

doi:10.1145/1989323.1989451

Cited by 64 publications

(61 citation statements)

References 23 publications

(14 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…To our knowledge, the Statistical Relational Models (SRMs) of Getoor, Taskar and Koller [8], are the only prior statistical models with a class-level probability semantics. A direct empirical comparison is di cult as code has not been released, but SRMs have compared favorably with benchmark methods for estimating the cardinality of a database query result [2] (Sec.…”

Section: Related Work and Discussionmentioning

confidence: 99%

“…(1) SRMs are derived from a tuple semantics [8,Def.6.3], which is different from the random selection semantics we propose for FBNs. (2) SRMs are less expressive: The queries that can be formulated using the nodes in an SRM cannot express general combinations of positive and negative relationships [8,Def.6.6]. This restriction stems from the fact that the SRM semantics is based on randomly selecting tuples from existing tables in the database.…”

Section: Related Work and Discussionmentioning

confidence: 99%

“…The ground graph can be used to answer instance-level queries about individuals, such as "if user sam has 3 friends, female rozita, males ali and victor, what is the probability that sam is a woman"? However, as Getoor pointed out [8], the ground graph is not appropriate for answering class-level queries, which are about generic rates and percentages, not about any particular individuals.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Modelling relational statistics with Bayes Nets

et al. 2013

View full text Add to dashboard Cite

Abstract. Class-level models capture relational statistics over object attributes and their connecting links, answering questions such as "what is the percentage of friendship pairs where both friends are women?" Class-level relationships are important in themselves, and they support applications like policy making, strategic planning, and query optimization. We represent class statistics using Parametrized Bayes Nets (PBNs), a first-order logic extension of Bayes nets. Queries about classes require a new semantics for PBNs, as the standard grounding semantics is only appropriate for answering queries about specific ground facts. We propose a novel random selection semantics for PBNs, which does not make reference to a ground model, and supports class-level queries. The parameters for this semantics can be learned using the recent pseudo-likelihood measure [1] as the objective function. This objective function is maximized by taking the empirical frequencies in the relational data as the parameter settings. We render the computation of these empirical frequencies tractable in the presence of negated relations by the inverse Möbius transform. Evaluation of our method on four benchmark datasets shows that maximum pseudo-likelihood provides fast and accurate estimates at different sample sizes.

show abstract

Section: Related Work and Discussionmentioning

confidence: 99%

Section: Related Work and Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Modelling relational statistics with Bayes Nets

et al. 2013

View full text Add to dashboard Cite

show abstract

“…One proposal [6] is designed with only selection queries in mind, and another [13] supports only key-foreign key joins. However, it is not hard to extend the notion of statistical relational models described in [12] to work correctly with arbitrary joins. The downside is that all possible joins must be known prior to building the statistical model.…”

Section: Estimation Of Joint Selectivitiesmentioning

confidence: 99%

Sharing-aware horizontal partitioning for exploiting correlations during query processing

2010

View full text Add to dashboard Cite

Optimization of join queries based on average selectivities is suboptimal in highly correlated databases. In such databases, relations are naturally divided into partitions, each partition having substantially different statistical characteristics. It is very compelling to discover such data partitions during query optimization and create multiple plans for a given query, one plan being optimal for a particular combination of data partitions. This scenario calls for the sharing of state among plans, so that common intermediate results are not recomputed. We study this problem in a setting with a routing-based query execution engine based on eddies [1]. Eddies naturally encapsulate horizontal partitioning and maximal state sharing across multiple plans. We define the notion of a conditional join plan, a novel representation of the search space that enables us to address the problem in a principled way. We present a lowoverhead greedy algorithm that uses statistical summaries based on graphical models. Experimental results suggest an order of magnitude faster execution time over traditional optimization for high correlations, while maintaining the same performance for low correlations.

show abstract

“…(See, for example, [3], [4], [9].) These collective classification techniques typically outperform classifiers that use only propositional data.…”

Section: Introductionmentioning

confidence: 99%

Multi-Stage Classification

Senator

Fifth IEEE International Conference on Data Mining (ICDM'05)

View full text Add to dashboard Cite

While much research has focused on methods for evaluating and maximizing the accuracy of classifiers either individually or in ensembles, little effort has been devoted to analyzing how classifiers are typically deployed in practice. In many domains, classifiers are used as part of a multi-stage process that increases accuracy at the expense of more data collection and/or more processing resources as the likelihood of a positive class label increases. This paper systematically explores the tradeoffs inherent in constructing these multi-stage classifiers from a series of increasingly accurate and expensive individual classifiers, considering a variety of metrics such as accuracy, cost/benefit ratio, and lift. It suggests architectures appropriate for both independent instances and for highly linked data.

show abstract

Learning statistical models from relational data

Cited by 64 publications

References 23 publications

Modelling relational statistics with Bayes Nets

Modelling relational statistics with Bayes Nets

Sharing-aware horizontal partitioning for exploiting correlations during query processing

Multi-Stage Classification

Contact Info

Product

Resources

About