Harris T. Lin

Learning relational bayesian classifiers from RDF data

The increasing availability of large RDF datasets offers an exciting opportunity to use such data to build predictive models using machine learning algorithms. However, the massive size and distributed nature of RDF data calls for approaches to learning from RDF data in a setting where the data can be accessed only through a query interface, e.g., the SPARQL endpoint of the RDF store. In applications where the data are subject to frequent updates, there is a need for algorithms that allow the predictive model to be incrementally updated in response to changes in the data. Furthermore, in some applications, the attributes that are relevant for specific prediction tasks are not known a priori and hence need to be discovered by the algorithm. We present an approach to learning Relational Bayesian Classifiers (RBCs) from RDF data that addresses such scenarios. Specifically, we show how to build RBCs from RDF data using statistical queries through the SPARQL endpoint of the RDF store. We compare the communication complexity of our algorithm with one that requires direct centralized access to the data and hence has to retrieve the entire RDF dataset from the remote location for processing. We establish the conditions under which the RBC models can be incrementally updated in response to addition or deletion of RDF data. We show how our approach can be extended to the setting where the attributes that are relevant for prediction are not known a priori, by selectively crawling the RDF data for attributes of interest. We provide open source implementation and evaluate the proposed approach on several large RDF datasets.

Triplet supertree heuristics for the tree of life

BackgroundThere is much interest in developing fast and accurate supertree methods to infer the tree of life. Supertree methods combine smaller input trees with overlapping sets of taxa to make a comprehensive phylogenetic tree that contains all of the taxa in the input trees. The intrinsically hard triplet supertree problem takes a collection of input species trees and seeks a species tree (supertree) that maximizes the number of triplet subtrees that it shares with the input trees. However, the utility of this supertree problem has been limited by a lack of efficient and effective heuristics.ResultsWe introduce fast hill-climbing heuristics for the triplet supertree problem that perform a step-wise search of the tree space, where each step is guided by an exact solution to an instance of a local search problem. To realize time efficient heuristics we designed the first nontrivial algorithms for two standard search problems, which greatly improve on the time complexity to the best known (naïve) solutions by a factor of n and n2 (the number of taxa in the supertree). These algorithms enable large-scale supertree analyses based on the triplet supertree problem that were previously not possible. We implemented hill-climbing heuristics that are based on our new algorithms, and in analyses of two published supertree data sets, we demonstrate that our new heuristics outperform other standard supertree methods in maximizing the number of triplets shared with the input trees.ConclusionWith our new heuristics, the triplet supertree problem is now computationally more tractable for large-scale supertree analyses, and it provides a potentially more accurate alternative to existing supertree methods.

Consensus properties for the deep coalescence problem and their application for scalable tree search

BackgroundTo infer a species phylogeny from unlinked genes, phylogenetic inference methods must confront the biological processes that create incongruence between gene trees and the species phylogeny. Intra-specific gene variation in ancestral species can result in deep coalescence, also known as incomplete lineage sorting, which creates incongruence between gene trees and the species tree. One approach to account for deep coalescence in phylogenetic analyses is the deep coalescence problem, which takes a collection of gene trees and seeks the species tree that implies the fewest deep coalescence events. Although this approach is promising for phylogenetics, the consensus properties of this problem are mostly unknown and analyses of large data sets may be computationally prohibitive.ResultsWe prove that the deep coalescence consensus tree problem satisfies the highly desirable Pareto property for clusters (clades). That is, in all instances, each cluster that is present in all of the input gene trees, called a consensus cluster, will also be found in every optimal solution. Moreover, we introduce a new divide and conquer method for the deep coalescence problem based on the Pareto property. This method refines the strict consensus of the input gene trees, thereby, in practice, often greatly reducing the complexity of the tree search and guaranteeing that the estimated species tree will satisfy the Pareto property.ConclusionsAnalyses of both simulated and empirical data sets demonstrate that the divide and conquer method can greatly improve upon the speed of heuristics that do not consider the Pareto consensus property, while also guaranteeing that the proposed solution fulfills the Pareto property. The divide and conquer method extends the utility of the deep coalescence problem to data sets with enormous numbers of taxa.

Clustering remote RDF data using SPARQL update queries

The emergence of large and distributed RDF data in the Linked Open Data cloud calls for approaches to extract useful knowledge using machine learning techniques such as clustering. However, the massive size and remote nature of RDF data hinder traditional approaches that gather the datasets onto a centralized location for analysis. In this work, we show how to implement two representative clustering algorithms using update queries against the SPARQL endpoint of the RDF store. We compare the time complexity and the communication complexity of our algorithms with of those that require direct centralized access to the data and hence have to retrieve the entire RDF dataset from the remote location. We conduct experiments on a real social network dataset and report our preliminary findings.

Learning Classifiers from Chains of Multiple Interlinked RDF Data Stores

The emergence of many interlinked, physically distributed, and autonomously maintained RDF stores offers unprecedented opportunities for predictive modeling and knowledge discovery from such data. However existing machine learning approaches are limited in their applicability because it is neither desirable nor feasible to gather all of the data in a centralized location for analysis due to access, memory, bandwidth, computational restrictions, and sometimes privacy and confidentiality constraints. Against this background, we consider the problem of learning predictive models from multiple interlinked RDF stores. Specifically we: (i) introduce statistical query based formulations of several representative algorithms for learning classifiers from RDF data, (ii) introduce a distributed learning framework to learn classifiers from multiple interlinked RDF stores that form a chain, (iii) identify three special cases of RDF data fragmentation and describe effective strategies for learning predictive models in each case, (iv) consider a novel application of a matrix reconstruction technique from the field of Computerized Tomography [1] to approximate the statistics needed by the learning algorithm from projections using count queries, thus dramatically reducing the amount of information transmitted from the remote data sources to the learner, and (v) report results of experiments with a real-world social network data set (Last.fm), which demonstrate the feasibility of the proposed approach.

The deep coalescence consensus tree problem is Pareto on clusters

Phylogenetic methods must account for the biological processes that create incongruence between gene trees and the species phylogeny. Deep coalescence, or incomplete lineage sorting creates discord among gene trees at the early stages of species divergence or in cases when the time between speciation events was short and the ancestral population sizes were large. The deep coalescence problem takes a collection of gene trees and seeks the species tree that implies the fewest deep coalescence events, or the smallest deep coalescence reconciliation cost. Although this approach can to be useful for phylogenetics, the consensus properties of this problem are largely uncharacterized, and the accuracy of heuristics is untested. We prove that the deep coalescence consensus tree problem satisfies the highly desirable Pareto property for clusters (clades). That is, in all instances, each cluster that is present in all of the input gene trees, called a consensus cluster, will also be found in every optimal solution. We introduce an efficient algorithm that, given a candidate species tree that does not display the consensus clusters, will modify the candidate tree so that it includes all of the clusters and has a lower (more optimal) deep coalescence cost. Simulation experiments demonstrate the efficacy of this algorithm, but they also indicate that even with large trees, most solutions returned by the recent efficient heuristic display the consensus clusters.

Learning Classifiers from Distributional Data

Many big data applications give rise to distributional data wherein objects or individuals are naturally represented as K-tuples of bags of feature values where feature values in each bag are sampled from a feature and object specific distribution. We formulate and solve the problem of learning classifiers from distributional data. We consider three classes of methods for learning distributional classifiers: (i) those that rely on aggregation to encode distributional data into tuples of attribute values, i.e., instances that can be handled by traditional supervised machine learning algorithms, (ii) those that are based on generative models of distributional data, and (iii) the discriminative counterparts of the generative models considered in (ii) above. We compare the performance of the different algorithms on real-world as well as synthetic distributional data sets. The results of our experiments demonstrate that classifiers that take advantage of the information available in the distributional instance representation outperform or match the performance of those that fail to fully exploit such information.

Learning classifiers from remote RDF data stores augmented with RDFS subclass hierarchies

Rapid growth of RDF data in the Linked Open Data (LOD) cloud offers unprecedented opportunities for analyzing such data using machine learning algorithms. The massive size and distributed nature of LOD cloud present a challenging machine learning problem where the data can only be accessed remotely, i.e. through a query interface such as the SPARQL end-point of the data store. Existing approaches to learning classifiers from RDF data in such a setting fail to take advantage of RDF schema (RDFS) associated with the data store that asserts subclass hierarchies which provide information that can potentially be exploited by the learner. Against this background, we present a general approach that augments an existing directed graphical model with hidden variables that encode subclass hierarchies via probabilistic constraints. We also present an algorithm ProbAVT that adopts the variational Bayesian expectation maximization approach to efficiently learn parameters in such settings. Our experiments with several synthetic and real world datasets show that: (i) ProbAVT matches or outperforms its counterpart that does not incorporate background knowledge in the form of subclass hierarchies; (ii) ProbAVT remains competitive compared to other state-of-art models that incorporate subclass hierarchies, and is able to scale up to large hierarchies consisting of over tens of thousands of nodes.