Knowledge-based Model Construction Research Articles

Bayesian Inference with Complex Knowledge Graph Evidence

Knowledge Graphs (KGs) provide a widely used format for representing entities and their relationships and have found use in diverse applications including question answering and recommendation. A majority of current research on KG inference has focused on reasoning with atomic facts (triples) and has disregarded the possibility of making complex evidential observations involving logical operators (negation, conjunction, disjunction) and quantifiers (existential, universal). Further, while the application of complex evidence has been explored in KG-based query answering (KGQA) research, in many practical online settings, observations are made sequentially. For example, in KGQA, additional context may be incrementally suggested to narrow down the answer. Or in interactive recommendation, user critiques may be expressed sequentially in order to narrow down a set of preferred items. Both settings are indicative of information filtering or tracking tasks that are reminiscent of belief tracking in Bayesian inference. In fact, in this paper, we precisely cast the problem of belief tracking over unknown KG entities given incremental complex KG evidence as a Bayesian filtering problem. Specifically, we leverage Knowledge-based Model Construction (KBMC) over the logical KG evidence to instantiate a Markov Random Field (MRF) likelihood representation to perform closed-form Bayesian inference with complex KG evidence (BIKG). We experimentally evaluate BIKG in incremental KGQA and interactive recommendation tasks demonstrating that it outperforms non-incremental methodologies and leads to better incorporation of conjunctive evidence vs. existing complex KGQA methods like CQD that leverage fuzzy T-norm operators. Overall, this work demonstrates a novel, efficient, and unified perspective of logic, KGs, and online inference through the lens of closed-form BIKG.

Computational state space models for activity and intention recognition. A feasibility study.

BackgroundComputational state space models (CSSMs) enable the knowledge-based construction of Bayesian filters for recognizing intentions and reconstructing activities of human protagonists in application domains such as smart environments, assisted living, or security. Computational, i. e., algorithmic, representations allow the construction of increasingly complex human behaviour models. However, the symbolic models used in CSSMs potentially suffer from combinatorial explosion, rendering inference intractable outside of the limited experimental settings investigated in present research. The objective of this study was to obtain data on the feasibility of CSSM-based inference in domains of realistic complexity.MethodsA typical instrumental activity of daily living was used as a trial scenario. As primary sensor modality, wearable inertial measurement units were employed. The results achievable by CSSM methods were evaluated by comparison with those obtained from established training-based methods (hidden Markov models, HMMs) using Wilcoxon signed rank tests. The influence of modeling factors on CSSM performance was analyzed via repeated measures analysis of variance.ResultsThe symbolic domain model was found to have more than states, exceeding the complexity of models considered in previous research by at least three orders of magnitude. Nevertheless, if factors and procedures governing the inference process were suitably chosen, CSSMs outperformed HMMs. Specifically, inference methods used in previous studies (particle filters) were found to perform substantially inferior in comparison to a marginal filtering procedure.ConclusionsOur results suggest that the combinatorial explosion caused by rich CSSM models does not inevitably lead to intractable inference or inferior performance. This means that the potential benefits of CSSM models (knowledge-based model construction, model reusability, reduced need for training data) are available without performance penalty. However, our results also show that research on CSSMs needs to consider sufficiently complex domains in order to understand the effects of design decisions such as choice of heuristics or inference procedure on performance.

Automated compilation of Object-Oriented Probabilistic Relational Models

This paper addresses the issues of knowledge representation and reasoning in large, complex, uncertain domains, focusing on the tactical military domain, which is characterized by all these properties. The key to reasoning efficiently under these circumstances is to provide a knowledge representation language and reasoning techniques which take advantage of the structure of the domain and facilitate reuse. First order representations such as relational logic are useful for representing structured domains because they can represent both entities and relations. However, the constraint that first order logic statements must be either true or false makes these languages unsuited to representing real world domains which involve uncertainty. Probability theory, on the other hand, provides a sound mathematical basis for representing and reasoning with uncertain information. For example, Bayesian Networks (BNs) are a well known probabilistic representation technique. However, there are several characteristics of large, complex domains which are challenging for traditional BNs. Recent research, for example [1,2], has shown there are advantages to be derived from combining probability theory with some of the expressive power of first order logics. Languages which combine probability theory with aspects of first order logic are called First Order Probabilistic Languages (FOPLs). There are a number of such languages, for example [1–4]. FOPLs have been used to model a number of domains such as military situation awareness [2], traffic surveillance [3], information extraction [5], natural language processing [6], intelligent tutoring systems [7], web site user behavior modelling [8] and automated internet fault diagnosis [9]. In this paper, we present the Object-Oriented Probabilistic Relational Modelling Language (OPRML) [10,11], a new FOPL which combines the generality and modularity of relational logic representation with a principled treatment of uncertainty. We describe the language in detail, outlining its formal syntax and semantics and compare it against its most closely related language: Probabilistic Relational Models (PRMs). We also present four novel algorithms for the automatic construction of domain models from knowledge-bases expressed using the OPRML. Two of the algorithms are based on the knowledge-based model construction approach and two are based on an Object-Oriented Bayesian Network instance tree triangulation method. We discuss the strengths and limitations of each of the algorithms and compare their performance against the algorithms developed for PRMs.

Knowledge-based time-critical dynamic decision modelling

Recent interests in dynamic decision modelling have led to the development of several representation and inference methods. These methods however, have limited application under time-critical conditions where a trade-off between model quality and computational tractability is essential. This paper presents a knowledge-based approach to time-critical dynamic decision modelling. A knowledge representation and modelling method, called the time-critical dynamic influence diagram, is proposed. The proposed approach has the ability to represent space-temporal abstraction in dynamic decision models. Several algorithms from different classes for solving time-critical dynamic influence diagrams are described. A knowledge-based meta-reasoning approach is proposed for the purpose of selecting the best abstracted model and algorithm that provide the optimal trade-off between model quality and model tractability. The approach is applied to solve a time-critical medical decision problem. An outline of the knowledge-based model construction procedure is also provided.

On the complexity of inference about probabilistic relational models

We investigate the complexity of probabilistic inference from knowledge bases that encode probability distributions on finite domain relational structures. Our interest here lies in the complexity in terms of the domain under consideration in a specific application instance. We obtain the result that assuming NETIME≠ ETIME this problem is not polynomial for reasonably expressive representation systems. The main consequence of this result is that it is unlikely to find inference techniques with a better worst-case behavior than the commonly employed strategy of constructing standard Bayesian networks over ground atoms (knowledge based model construction).

An Overview of Some Recent Developments in Bayesian Problem-Solving Techniques

The last few years have seen a surge in interest in the use of techniques from Bayesian decision theory to address problems in AI. Decision theory provides a normative framework for representing and reasoning about decision problems under uncertainty. Within the context of this framework, researchers in uncertainty in the AI community have been developing computational techniques for building rational agents and representations suited to engineering their knowledge bases. This special issue reviews recent research in Bayesian problem-solving techniques. The articles cover the topics of inference in Bayesian networks, decision-theoretic planning, and qualitative decision theory. Here, I provide a brief introduction to Bayesian networks and then cover applications of Bayesian problem-solving techniques, knowledge-based model construction and structured representations, and the learning of graphic probability models.

An Architecture for Knowledge-based Construction of Decision Models

Clinical application of decision analysis has been limited by unfamiliarity of clinicians with the technique, large data requirements, and the length of time needed to construct models. In order to make decision modeling more accessible to clinicians, the authors developed a computer program to construct decision models automatically. The system contains two separate knowledge bases. One contains frames encoding knowledge of the medical domain, the evaluation of pulmonary disease in patients infected with the human immunodeficiency virus (HIV). The other contains rules of correct decision model construction that guide the selection of items from the domain knowledge base and their insertion into the decision model. The system can create either a tree or an influence diagram that satisfies previously published critiquing rules. The system has the potential to enable novices to construct useful decision models and to provide individualized decision-analytic advice to clinicians in real time.

Computer-aided model construction

We examine ways in which the construction of mathematical models may be supported, and review several approaches for computer-aided model construction. The construction of complex models can be a challenging task even for expert modelers. The aim of computer-aided model construction systems is to simplify this task. We view model construction as a state transformation process, and suggest that the process can be facilitated by supporting the creation of certain representations or by transforming an existing representation to another. For example, declarative modeling languages facilitate the specification of an executable statement of the mathematical model, while knowledge-based model construction systems assist the creation of the mathematical formulation. Knowledge-based model construction systems are the focus of our review of existing systems. To facilitate a comparison of several such systems, we characterize them in terms of their cognitive bases, the representations developed in them, and the methods for transformation of these representations.

Knowledge-based model construction for simulating information systems

The process of constructing a correct simulation model for a system is a difficult and time-consuming task in the simulation life-cycle. Simulation is a widely used methodology for checking the performance requirements of information systems. This paper describes a knowledge-based approach to constructing a simulation model of an information system. The approach consists of two major steps: first, the information system is modeled as a Petri Net and implemented in a logic-based language. Second, the executable representation of the Petri Net is used to automatically construct a simulation model of the information system in GPSS. The approach offers several benefits such as reducing system development time, allowing the flexibility of easy modifications to the information system configuration, and requiring little knowledge of simulation methodology. The knowledge-based system is implemented in PROLOG.

Knowledge-based simulation model specification

A Knowledge-Based Model Construction (KBMC) system has been developed to automate the model construction phase of the simulation life cycle. The system's underlying rule base in corporates several types of knowledge. This includes domain knowledge that facilitates a structured interactive dialog for the acquisition of a complete model specification from a user. An executable discrete simulation model in SIMAN is automatically constructed by the system from this specification, utilizing model ing knowledge and SIMAN knowledge. We present an overview of the KBMC system and focus on the knowledge-based specifica tion method used in the KBMC system.