Probabilistic conditional logics offer a rich and well-founded framework for designing expert systems. The factorization of their maximum entropy models has several interesting applications. In this paper a general factorization is derived providing a more rigorous proof than in previous work. It yields an approach to extend Iterative Scaling variants to deterministic knowledge bases. Subsequently the connection to Markov Random Fields is revisited.

Situations where there is insufficient information to learn from often arise, and the process to recollect data can be expensive or in some cases take too long resulting in outdated models. Transfer learning strategies have proven to be a powerful technique to learn models from several sources when a single source does not provide enough information. In this work we present a methodology to learn a Temporal Nodes Bayesian Network by transferring knowledge from several different but related domains. Experiments based on a reference network show promising results, supporting our claim that transfer learning is a viable strategy to learn these models when scarce data is available.

Any computing with words (CW) system is required to assign a phrase in natural language to the fuzzy values it provides as its output. This paper explores different linguistic approximation methods for CW systems. The outputs of these methods are evaluated through various measures such as fuzziness, specificity, validity, and sigma-count. We illustrate that certain linguistic methods may result in complex and incomprehensible phrases in natural language. Some might even include an invalid linguistic term in their linguistic approximation.

The availability of massive amounts of raw domain data has created an urgent need for sophisticated AI systems with capabilities to find complex and useful information in big-data repositories in real-time. Such systems should have capabilities to process and extract significant information from natural language documents, search and answer complex questions, make sophisticated predictions about future events, and generally interact with users in much more powerful and intuitive ways. To be effective, these systems need a significant amount of domain-specific knowledge in addition to the general-domain knowledge. Ontologies/Knowledge-Bases represent knowledge about domains of interest and serve as the backbone for semantic technologies and applications. However, creating such domain models is time consuming, error prone, and the end product is difficult to maintain. In this paper, we present a novel methodology to automatically build semantically rich knowledge models for specific domains using domain-relevant unstructured data from resources such as web articles, manuals, e-books, blogs, etc. We also present evaluation results for our automatic ontology/knowledge-base generation methodology using freely-available textual resources from the World Wide Web.

The Dominance-based Rough Set Approach (DRSA) is an extension of Rough Sets Theory to handle multicriteria classification problems by authorizing preference-ordered attributes. The DRSA assumes the existence of a single decision table while real-world decision problems imply generally several experts with different decision tables. The objective of this paper is to propose an algorithm for the aggregation of a set of decision tables, as a first step for approximating these tables. The algorithm is illustrated using real-world data.

We compare three approaches to learning numerical parameters of Bayesian networks from continuous data streams: (1) the EM algorithm applied to all data, (2) the EM algorithm applied to data increments, and (3) the online EM algorithm. Our results show that learning from all data at each step, whenever feasible, leads to the highest parameter accuracy and model classification accuracy. When facing computational limitations, incremental learning approaches are a reasonable alternative. Of these, online EM is reasonably fast, and similar to the incremental EM algorithm in terms of accuracy. For small data sets, incremental EM seems to lead to better accuracy. When the data size gets large, online EM tends to be more accurate.

In the domain of decision theoretic planning, the factored framework (FMDP) has produced optimized algorithms using Decision Trees (SVI, SPI) and Algebraic Decision Diagrams (SPUDD). However, the state-of-the-art SPUDD algorithm requires i) the problem to be specified with binary variables and ii) the data structures to share a common order on variables. In this article, we propose a new algorithm within the factored framework that eliminates both these requirements. We compare our approach to the SPUDD algorithm. Experimental results show that our algorithm allows significant gains in time, illustrating a better trade-off between theoretical complexity of algorithms and size of representation.

Similar to Bayesian networks, so-called OCF-networks combine structural information encoded in a directed graph with qualitative information expressed by ranking degrees of (conditional) formulas. The benefits of such techniques are twofold: First, the high complexity of the semantical ranking functions approach is reduced substantially, and second, global ranks are obtained from local information. However, in many practical applications, even the local rankings are only available in parts, or not exactly in the format that is needed. In this paper, we apply inductive reasoning methods like system Z+ or c-representations, to fill up missing values in the local conditional tables. This allows the user to specify knowledge for such OCF-networks in its most appropriate and reliable form and leave the technical details to an inference engine.

Comparative preference statements are the basic ingredients of conditional logics for representing users’ preferences in a compact way. These statements may be strict or not and obey different semantics. Algorithms have been developed in the literature to compute a preference relation over outcomes given a set of comparative preference statements and one or several semantics. These algorithms are based on insights from non-monotonic reasoning (more specifically, minimal and maximal specificity principles) enforcing the preference relations to be a complete preorder. The main limitation of these logics however relies in preference queries when comparing two outcomes. Indeed given two outcomes having the same preference w.r.t. the preference relation, there is no indication whether this equality results from an equality between two preference statements or the outcomes are in fact incomparable and equality has been enforced by specificity principles. On the other hand, comparative preference statements and their associated semantics can be translated into qualitative constraint satisfaction problems in which one can have a precise ordering over two outcomes. In this paper we investigate this bridge and provide a compilation of conditional logics-based preference queries in qualitative constraint problems.

We present a new approach to reasoning about the outcome of an argumentation framework, where an agent's reasoning with a framework and semantics is represented by an inference relation defined over a logical labeling language. We first study a monotonic type of inference which is, in a sense, more general than an acceptance function, but equally expressive. In order to overcome the limitations of this expressiveness, we study a non-monotonic type of inference which allows counterfactual inferences. We precisely characterize the classes of frameworks distinguishable by the non-monotonic inference relation for the admissible semantics.