Search

Sum-Product Networks (SPNs) are recently introduced deep probabilistic models providing exact and tractable inference. SPNs have been successfully employed in several application domains, from computer vision to natural language processing, as accurate density estimators. However, learning their structure and parameters from high dimensional data poses a challenge in terms of time complexity. Classical SPNs structure learning algorithms work by repeating several times two high cost operations: determining independencies among random variables (RVs)–introducing product nodes–and finding sub-populations among samples–introducing sum nodes. Even one of the simplest greedy structure learner, LearnSPN , scales quadratically in the number of the variables to determine RVs independencies. In this work, we investigate the trade-off between accuracy and efficiency when employing approximate but fast procedures to determine independencies among RVs. We introduce and evaluate sub-quadratic procedures based on a random subspace approach and leveraging entropy as a proxy criterion to split independent RVs. Experimental results on many benchmark datasets for density estimation show that LearnSPN -like structure learners, when equipped by our splitting procedures, provide reduced learning and/or inference times, generally containing the degradation of inference accuracy. Ultimately, we provide an empirical confirmation of a "no free lunch" when learning the structure of SPNs. [ABSTRACT FROM AUTHOR]

Standardized processes are important for correctly carrying out activities in an organization. Often the procedures they describe are already in operation, and the need is to understand and formalize them in a model that can support their analysis, replication and enforcement. Manually building these models is complex, costly and error-prone. Hence, the interest in automatically learning them from examples of actual procedures. Desirable options are incrementality in learning and adapting the models, and the ability to express triggers and conditions on the tasks that make up the workflow. This paper proposes a framework based on First-Order Logic that solves many shortcomings of previous approaches to this problem in the literature, allowing to deal with complex domains in a powerful and flexible way. Indeed, First-Order Logic provides a single, comprehensive and expressive representation and manipulation environment for supporting all of the above requirements. A purposely devised experimental evaluation confirms the effectiveness and efficiency of the proposed solution. [ABSTRACT FROM AUTHOR]

This work focusses on the problem of clustering resources contained in knowledge bases represented throughmulti-relational standard languages that are typical for the context of the Semantic Web, and ultimately founded in Description Logics. The proposed solution relies on effective and language-independent dissimilarity measures that are based on a finite number of dimensions corresponding to a committee of discriminating features, that stands for a context, represented by concept descriptions in Description Logics. The proposed clustering algorithm expresses the possible clusterings in tuples of central elements: in this categorical setting, we resort to the notion of medoid, w.r.t. the given metric. These centers are iteratively adjusted following the rationale of fuzzy clustering approach, i.e. one where the membership to each cluster is not deterministic but graded, ranging in the unit interval. This better copes with the inherent uncertainty of the knowledge bases expressed in Description Logics which adopt an open-world semantics. An extensive experimentation with a number of ontologies proves the feasibility of our method and its effectiveness in terms of major clustering validity indices. [ABSTRACT FROM AUTHOR]

The issue addressed in this paper concerns the discovery of frequent multi-dimensional patterns from relational sequences. The great variety of applications of sequential pattern mining, such as user profiling, medicine, local weather forecast and bioinformatics, makes this problem one of the central topics in data mining. Nevertheless, sequential information may concern data on multiple dimensions and, hence, the mining of sequential patterns from multi-dimensional information results very important. In a multi-dimensional sequence each event depends on more than one dimension, such as in spatio-temporal sequences where an event may be spatially or temporally related to other events. In literature, the multi-relational data mining approach has been successfully applied to knowledge discovery fromcomplex data. However, there exists no contribution to manage the general case of multi-dimensional data in which, for example, spatial and temporal information may co-exist. This work takes into account the possibility to mine complex patterns, expressed in a first-order language, in which events may occur along different dimensions. Specifically, multidimensional patterns are defined as a set of atomic first-order formulae in which events are explicitly represented by a variable and the relations between events are represented by a set of dimensional predicates. A complete framework and an Inductive Logic Programming algorithm to tackle this problem are presented along with some experiments on artificial and real multi-dimensional sequences proving its effectiveness. [ABSTRACT FROM AUTHOR]

Traditional Machine Learning approaches based on single inference mechanisms have reached their limits. This causes the need for a framework that integrates approaches based on abduction and abstraction capabilities in the inductive learning paradigm, in the light of Michalski's Inferential Theory of Learning (ITL). This work is intended as a survey of the most significant contributions that are present in the literature, concerning single reasoning strategies and practical ways for bringing them together and making them cooperate in order to improve the effectiveness and efficiency of the learning process. The elicited role of an abductive proof procedure is tackling the problem of incomplete relevance in the incoming examples. Moreover, the employment of abstraction operators based on (direct and inverse) resolution to reduce the complexity of the learning problem is discussed. Lastly, a case study that implements the combined framework into a real multistrategy learning system is briefly presented. [ABSTRACT FROM AUTHOR]

Model trees are an extension of regression trees that associate leaves with multiple regression models. In this paper, a method for the data-driven construction of model trees is presented, namely, the Stepwise Model Tree Induction (SMOTI) method. Its main characteristic is the induction of trees with two types of nodes: regression nodes, which perform only straight-line regression, and splitting nodes, which partition the feature space. The multiple linear model associated with each leaf is then built stepwise by combining straight-line regressions reported along the path from the root to the leaf. In this way, internal regression nodes contribute to the definition of multiple models and have a "global" effect, while straight-line regressions at leaves have only "local" effects. Experimental results on artificially generated data sets show that SMOTI outperforms two model tree induction systems, M5' and RETIS, in accuracy. Results on benchmark data sets used for studies on both regression and model trees show that SMOTI performs better than RETIS in accuracy, while it is not possible to draw statistically significant conclusions on the comparison with M5'. Model trees induced by SMOTI are generally simple and easily interpretable and their analysis often reveals interesting patterns. [ABSTRACT FROM AUTHOR]

The data descriptions of the units are called "symbolic" when they are more complex than standard ones, due to the fact that they contain internal variations and are structured. Symbolic data arise from many sources, for instance when summarizing huge Relational Data Bases by their underlying concepts. "Extracting knowledge" means obtaining explanatory results, and for this reason, "symbolic objects" are introduced and studied in this paper. They model concepts and constitute an explanatory output for data analysis. Moreover, they can be used to define queries of a Relational Data Base and propagate concepts between Data Bases. We define "Symbolic Data Analysis" (SDA) as the extension of standard Data Analysis to symbolic data tables as input in order to find symbolic objects as output. Any SDA is based on four spaces: the space of individuals, the space of concepts, the space of descriptions modelling individuals or classes of individuals, the space of symbolic objects modelling concepts. New problems arise from these four spaces, such as the quality, robustness and reliability of the approximation of a concept given by a symbolic object, the symbolic description of a class, the consensus between symbolic descriptions, and so on. In this paper we give an overview of recent developments in SDA. We briefly describe some SDA tools and methods and, in particular, we describe some dissimilarity methods for symbolic objects which are central to the majority of symbolic data analysis methods. Finally, we introduce the software prototype, developed by 17 teams from nine countries involved in the SODAS EUROSTAT project. [ABSTRACT FROM AUTHOR]

This work presents the application of a multistrategy approach to some document processing tasks. The application is implemented in an enhanced version of the incremental learning system INTHELEX. This learning module has been embedded as a learning component in the system architecture of the EU project COLLATE, which deals with the annotation of cultural heritage documents. Indeed, the complex shape of the material handled in the project has suggested that the addition of multistrategy capabilities is needed to improve effectiveness and efficiency of the learning process. Results proving the benefits of these strategies in specific classfication tasks are reported in the experimentation presented in this work. [ABSTRACT FROM AUTHOR]

Information given in topographic map captions or in GIS models is often insufficient to recognize interesting geographical patterns. Some prototypes of GIS have already been extended with a knowledge-base and some reasoning capabilities to support sophisticated map interpretation processes. Nevertheless, the acquisition of the necessary knowledge is still a demanding task for which machine learning techniques can be of great help. This paper presents INGENS, a prototypical GIS which integrates machine learning tools to assist users in the task of topographic map interpretation. The system can be trained to learn operational definitions of geographical objects that are not explicitly modeled in the database. INGENS has been applied to the task of Apulian map interpretation in order to discover geographic knowledge of interest to town planners. [Copyright &y& Elsevier]

Numerical induction from data is one of the statistical data analysis tasks, which uses a tabular model, with almost exclusively numerical features, as data representation formalism. The output representations are different: from functions to probability distributions, from surface equations to tables of indexes. One approach to extend the classical data analysis techniques to symbolic objects is the Symbolic Data Analysis: the input and the output of classical techniques are expressed in a symbolic way, so guaranteeing the comprehensibility of both the observations and the results, while the processing techniques, although appropriately adapted, maintain the efficiency of the classical statistical inferential models. Also in the field of Machine Learning several methods have been proposed to extend some inductive approaches from statistical data analysis to data represented as attribute-value couples. Sometimes these approaches transform ideas and principles coming from numerical induction to handle propositional calculus descriptions, otherwise they combine different techniques in order to treat numerical-continuous data and algebraic-symbolic data differently. The aim here is to improve the efficiency and to preserve the expressive power of the representations during the learning process, and to save the accuracy and flexibility of the numerical techniques during the recognition phase. This kind of integration is more and more complex when using first-order computational learning models, which are useful for handling object descriptions in structured domains, when not only the properties of objects but also the relations between different objects must be considered. The necessity arises from integrating different computational strategies, different knowledge representations and different processing methods in a naive combination of classifiers or, more meaningfully, in a real integration within a unique theoretical framework. When building machine learn... [ABSTRACT FROM AUTHOR]

Editorial. Examines the areas of artificial intelligence. Details on the interaction between computer vision and machine learning; Task of computer vision systems; Aspects of image and signal processing.

Presents a study that developed a framework for theory refinement that fulfills properties deemed as desirable for the incremental inductive synthesis of logic-based knowledge. Information on the basic notions of the representation language; Properties required for substitutions in order to fulfill object identity; Discussion on refinement operators and their properties; Details of an ideal refinement operator for object identity-implication.

The design of a user interface integrating instruments for visual and textual representation and image interpretation is a relevant problem when developing an advisory system for environmental planning. Indeed, the user of the system needs a support to the interpretation of maps, that is, a tool that segments maps and automatically associates geometric regions on a map with those semantic labels useful for applying hints and advices suggested by the environmental planning system. In the article, we present the application of symbolic machine learning techniques to the interpretation of maps. Two inductive learning systems, namely, INDUBI/CSL and ATRE, have been used to complete the knowledge base of an expert system for environmental planning. The application described concerns the recognition of four environmental concepts that are relevant for environmental protection. The positive results obtained in two different experiments prove the strength of the adopted approach for the interpretation task. [ABSTRACT FROM AUTHOR]

This article explores the combined application of inductive learning algorithms and causal inference techniques to the problem of discovering causal rules among the attributes of a relational database. Given some relational data each field can be considered as a random variable and a hybrid graph can be built by detecting conditional independencies among variables. The induced graph represents genuine and potential causal relations as well as spurious associations. When the variables are discrete or have been discretized to test condi tional independencies supervised induction algorithms can be used to learn causal rules that is conditional statements in which causes appear as antecedents and effects as consequences. The approach is illustrated by means of some experiments conducted on different data sets. [ABSTRACT FROM AUTHOR]

Addresses the problem of retrospectively pruning decision trees induced from data, according to a top-down approach. Proposition of various methods to address problem; Comparative study of various pruning methods; Characteristics of each method.

Abstract. This paper presents IDL, a prototypical digital library service. It integrates machine learning tools and intelligent techniques in order to make effective, efficient and economically feasible the process of capturing the information that should be stored and indexed by content in the digital library. In fact, information capture and semantic indexing are critical issues when building a digital library, since they involve complex pattern recognition problems, such as document analysis, classification and understanding. Experimental results show that learning systems can effectively and efficiently solve all these problems. [ABSTRACT FROM AUTHOR]

Abstract: Metabolomics is increasingly becoming an important field. The fundamental task in this area is to measure and interpret complex time and condition dependent parameters such as the activity or flux of metabolites in cells, their concentration, tissues elements and other biosamples. The careful study of all these elements has led to important insights in the functioning of metabolism. Recently, however, there is a growing interest towards an integrated approach to studying biological systems. This is the main goal in Systems Biology where a combined investigation of several components of a biological system is thought to produce a thorough understanding of such systems. Biological circuits are complex to model and simulate and many efforts are being made to develop models that can handle their intrinsic complexity. A significant part of biological networks still remains unknown even though recent technological developments allow simultaneous acquisition of many metabolite measurements. Metabolic networks are not only structurally complex but behave also in a stochastic fashion. Therefore, it is necessary to express structure and handle uncertainty to construct complete dynamics of these networks. In this paper we describe how stochastic modeling and simulation can be performed in a symbolic-statistical machine learning (ML) framework. We show that symbolic ML deal with structural and relational complexity while statistical ML provides principled approaches to uncertainty modeling. Learning is used to analyze traces of biochemical reactions and model the dynamicity through parameter learning, while inference is used to produce stochastic simulation of the network. [Copyright &y& Elsevier]

Abstract: A conceptual clustering framework is presented which can be applied to multi-relational knowledge bases storing resource annotations expressed in the standard languages for the Semantic Web. The framework adopts an effective and language-independent family of semi-distance measures defined for the space of individual resources. These measures are based on a finite number of dimensions corresponding to a committee of discriminating features represented by concept descriptions. The clustering algorithm expresses the possible clusterings in terms of strings of central elements (medoids, w.r.t. the given metric) of variable length. The method performs a stochastic search in the space of possible clusterings, exploiting a technique based on genetic programming. Besides, the number of clusters is not necessarily required as a parameter: a natural number of clusters is autonomously determined, since the search spans a space of strings of different length. An experimentation with real ontologies proves the feasibility of the clustering method and its effectiveness in terms of standard validity indices. The framework is completed by a successive phase, where a newly constructed intensional definition, expressed in the adopted concept language, can be assigned to each cluster. Finally, two possible extensions are proposed. One allows the induction of hierarchies of clusters. The other applies clustering to concept drift and novelty detection in the context of ontologies. [Copyright &y& Elsevier]

Symbolic data analysis aims at generalizing some standard statistical data mining methods, such as those developed for classification tasks, to the case of symbolic objects (SOs). These objects synthesize information concerning a group of individuals of a population, eventually stored in a relational database, and ensure confidentiality of original data. Classifying SOs is an important task in symbolic data analysis. In this paper a lazy-learning approach that extends a traditional distance weighted k-Nearest Neighbor classification algorithm to SOs, is presented. The proposed method has been implemented in the system SO-NN (Symbolic Objects Nearest Neighbor) and evaluated on symbolic datasets. [ABSTRACT FROM AUTHOR]

In Artificial Intelligence with Coalition Structure Generation (CSG) one refers to those cooperative complex problems that require to find an optimal partition (maximizing a social welfare) of a set of entities involved in a system. The solution of the CSG problem finds applications in many fields such as Machine Learning (set covering machines, clustering), Data Mining (decision tree, discretization), Graph Theory, Natural Language Processing (aggregation), Semantic Web (service composition), and Bioinformatics. The problem of finding the optimal coalition structure is NP-complete. In this paper we present a greedy adaptive search procedure (GRASP) with path-relinking to efficiently search the space of coalition structures. Experiments and comparisons to other algorithms prove the validity of the proposed method in solving this hard combinatorial problem. [ABSTRACT FROM AUTHOR]

With the increasing amount of information in electronic form the fields of Machine Learning and Data Mining continue to grow by providing new advances in theory, applications and systems. The aim of this paper is to consider some recent theoretical aspects and approaches to ML and DM with an emphasis on the Italian research. [ABSTRACT FROM AUTHOR]