Detection of anomalies (i.e., outliers ) in multi-dimensional data is a well-studied subject in machine learning. Unfortunately, un-supervised detectors provide no explanation about why a data point was considered as abnormal or which of its features (i.e. subspaces) exhibit at best its outlyingness. Such outlier explanations are crucial to diagnose the root cause of data anomalies and enable corrective actions to prevent or remedy their effect in downstream data processing. In this work, we present a comprehensive framework for comparing different unsupervised outlier explanation algorithms that are domain and detector-agnostic. Using real and synthetic datasets, we assess the effectiveness and efficiency of two point explanation algorithms (Beam [28] and RefOut [18]) ranking subspaces that best explain the outlyingness of individual data points and two explanation summarization algorithms (LookOut [15] and HiCS [17]) ranking subspaces that best exhibit as many outlier points from inliers as possible. To the best of our knowledge, this is the first detailed evaluation of existing explanation algorithms aiming to uncover several missing insights from the literature such as: (a) Is it effective to combine any explanation algorithm with any off-the-shelf outlier detector? (b) How is the behavior of an outlier detection and explanation pipeline affected by the number or the correlation of features in a dataset? and (c) What is the quality of summaries in the presence of outliers explained by subspaces of different dimensionality?

The goal of operational Business Intelligence (BI) is to help organizations improve the efficiency of their business by giving every "operational worker" insights needed to make better operational decisions, and aligning day-to-day operations with strategic goals. Operational BI reporting contributes to this goal by embedding analytics and reporting information into workflow applications so that the business user has all required information (contextual and business data) in order to make good decisions. EII systems facilitate the construction of operational BI reports by enabling the creation and querying of customized virtual database schemas over a set of distributed and heterogeneous data sources with a low TCO. Queries over these virtual databases feed the operational BI reports. We describe the characteristics of operational BI reporting applications and show that they increase the complexity of the source to target mapping defined between source data and virtual databases. We show that this complexity yields the execution of "mega queries", i.e., queries with possible a 1,000 tables in their FROM clause. We present some key optimization methods that have been successfully implemented in SAP Business Objects Data Federator system to deal with mega queries.

The goal of operational Business Intelligence (BI) is to help organizations improve the efficiency of their business by giving every "operational worker" insights needed to make better operational decisions, and aligning day-to-day operations with strategic goals. Operational BI reporting contributes to this goal by embedding analytics and reporting information into workflow applications so that the business user has all required information (contextual and business data) in order to make good decisions. EII systems facilitate the construction of operational BI reports by enabling the creation and querying of customized virtual database schemas over a set of distributed and heterogeneous data sources with a low TCO. Queries over these virtual databases feed the operational BI reports. We describe the characteristics of operational BI reporting applications and show that they increase the complexity of the source to target mapping defined between source data and virtual databases. We show that this complexity yields the execution of "mega queries", i.e., queries with possible a 1,000 tables in their FROM clause. We present some key optimization methods that have been successfully implemented in SAP Business Objects Data Federator system to deal with mega queries.

An active database system automatically triggers rules in response to certain events occuring. Events are issued by transactions or action parts of rules. Repeated executions of rules can be caused by the structure of the initial triggering transaction program and by the structure and execution semantics of rules. Repeated calculations of rules may incur costly redundant computations in rule conditions or actions. The central contribution of this paper is to propose technics for analyzing the behaviour of a transaction and a set of rules triggered by this transaction in order to derive: (i) if a given rule is processed more than once, and (ii) a ne indication of the database changes that may occur between two consecutive executions of the rule. Knowing these changes, it is possible to use existing algorithms that compute useful intermediate expressions in a rule that can be cached and incrementally maintained in order to avoid redundant computations. A notable property of our analysis technics is that they are parametrized by a few essential semantics parameters that deene the execution semantics of an active rule language. Thus, our analysis apply to a large class of existing active rule systems.

There exist many documents outlining methodologies for Systems Development Life Cycles. Although these documents are well adapted for human beings to follow, there is no system supporting the inherent complexity of the involved processes. In this paper, we present scalable social protocols, a recursive extension of social protocols, a formalism for representing human-to-human interactions. We argue that such a recursive definition allows working on a representation of the existing standards at different levels of detail. The resulting flexibility is essential to model real and complex systems and, ultimately, to implement dynamic and highly specialized life cycles.

The NeuroBase project aims at studying the requirements for federating, through the Internet, information sources in neuroimaging. These sources are distributed in different experimental sites, hospitals or research centers in cognitive neurosciences, and contain heterogeneous data and image processing programs. More precisely, this project consists in creating of a shared ontology, suitable for supporting various neuroimaging applications, and a computer architecture for accessing and sharing relevant distributed information. We briefly describe the semantic model and report in more details the architecture we chose, based on a media-tor/wrapper approach. To give a flavor of the future deployment of our architecture, we de-scribe a demonstrator that implements the comparison of distributed image processing tools applied to distributed neuroimaging data

Research in neuroscience makes an extensive use of modern neuroimaging data. The ability to share such data as well as processing tools to analyze them becomes a key factor of success of future research. This paper reports about works carried out in the context of the Neurobase project, a collaborative exploratory action supported by the French Ministry of Research. The partners have studied the design of a system suitable for sharing heterogeneous data and image processing tools, according to a federated approach, based on a common ontology. A demonstrator has been implemented, providing basic services for querying data and launching processing tools distributed at several sites.

This paper introduces a model that combines and extends capabilities of relational and object oriented systems. Within a simple, statically-typed framework, the model supports abstract data types (ADTs), a multiple inheritance ordering on ADTs based on behavioral similarity, multi-targeted generic operations, polymorphic methods, type-independent persistence, and set-oriented rules over shared objects, values, classes, and relations. The model maintains a clear separation between interface and implementation, and resolves currently open problems for typed OODBs.

We address the problem of efficient evaluation of non-deterministic logic programs in a database context using incremental evaluation algorithms. We consider a representative non-deterministic Datalog extension which gives a formal basis to various production rule languages. We point out a clear space-time tradeoff in the choice made by some algorithms to materialize and incrementally maintain some data in order to compute the meaning of a rule program. We advocate that the data maintained by an incremental evaluation algorithm should be chosen according to some particular properties of a rule program. We identify such properties and provide means to determine them statically based on a syntactic analysis of the bodies and heads of rules. We show that using these properties, one can improve the performance of the two best known incremental evaluation algorithms TREAT and RETE.

Active rules or triggers are widely accepted as powerful mechanisms to implement applications or systems behaviour. Several rule execution models were proposed as extended functionalities for different database systems. However, these models lack in flexibility and adaptability to specific database systems or specific application requirements. In this paper, we propose a generic framework which provides a set of basic functions which can be used to implement any execution model. This framework, called Active-Design toolkit, can be exploited in many situations where applications need specific model or different models for different subsets of rules.

In this paper, we consider the problem of integrating a production rule language, named RDL1, with a relational DBMS. A production rule in RDL1, consists of a condition part which is a relational calculus expression and of an action part which is a sequence of database updates. The main problem addressed in this paper is to determine whether a rule program can be computed as a relational algebra program, i.e., whether the initial semantics of the program is not modified by a set-oriented or relational computation. First, we define the syntax and the semantics of the RDL1 language which is given as the sequence of database states reachable by the computation of the program. We conjecture that deciding if a rule is relational computable is an undecidable problem and then, propose sufficient conditions to decide if a rule is relational computable. We present a general method to check the validity of these conditions. Finally, we propose two algorithms which are derived from the previous method. The first one gave sufficient syntactic conditions for a rule to be relational computable. The second one gave sufficient semantic conditions and leads to check integrity constraints over the database to decide whether a rule is relational computable.

The Neurobase project has for objective to establish the conditions allowing the federation through Internet of information sources in neuroimaging, where sources are distributed in different experimental sites, hospitals or research centers in cognitive neurosciences and contain data and image processing methods. This project consists in the definition of a computer architecture allowing the access and the sharing of such heterogeneous and autonomous sources in neuroimaging. This will enable for example, within these information sources, the search for similar data, the search for images containing singularities, the construction of a specific chain of distributed image processing tools or the transverse search of data for excavating possible regularities (similarly to a data mining approach).