Search

A plethora of machine learning and deep learning methods are used for the assessment of Alzheimer’s Disease (AD) from brain structural changes as seen in Magnetic Resonance Imaging (MRI) with highly satisfactory results. However, these models are black-box and lack an explicit declarative knowledge representation and thus there is a difficulty in generating the underlying explanatory imaging structures. The objective of this study was to investigate the usefulness of rule extraction in the assessment of AD using decision trees (DT) and random forests (RF) algorithms and integrating the extracted rules within an argumentation-based reasoning framework in order to make the results easy to interpret and explain. The DT and RF algorithms were applied on brain MRI images acquired from normal controls (NC) and AD subjects. The KNIME analytics platform was used to compute the DT and the R project was used for the RF. The argumentation model implemented in the Gorgias framework achieved an average accuracy of 91%, exhibiting improved results compared to the models of DT and RF. The overall performance of all models in this study is in agreement with other studies. In addition, the explanations given by our approach for the various possible predictions provide a more useful and complete assessment of the state of the patient/case at hand. This study demonstrated the usefulness of rule extraction in the assessment of AD based on MRI features and the positive results of the use of the argumentation based symbolic reasoning for composing and interpreting the ML results.

We describe Modular-E (ME), a specialized, model-theoretic logic for reasoning about actions. ME is able to represent non-deterministic domains involving concurrency, static laws (constraints), indirect effects (ramifications), and narrative information in the form of action occurrences and observations along a time line. We give formal results which characterize ME's high degree of modularity and elaboration tolerance, and show how these properties help to separate out, and provide principled solutions to, different aspects of the qualification problem. In particular, we identify the endogenous qualification problem as the problem of properly accounting for highly distributed, and potentially conflicting, causal knowledge when reasoning about the effects of actions. We show how a comprehensive solution to the endogenous qualification problem helps simplify the exogenous qualification problem - the problem of reconciling conflicts between predictions about what should be true at particular times and actual observations. More precisely, we describe how ME is able to use straightforward default reasoning techniques to solve the exogenous qualification problem largely because its robust treatments of the frame, ramification and endogenous qualification problems combine into a particular characteristic of elaboration tolerance that we formally encapsulate as a notion of ''free will''.

This paper presents the computational logic foundations of a model of agency called the KGP (Knowledge, Goals and Plan model. This model allows the specification of heterogeneous agents that can interact with each other, and can exhibit both proactive and reactive behaviour allowing them to function in dynamic environments by adjusting their goals and plans when changes happen in such environments. KGP provides a highly modular agent architecture that integrates a collection of reasoning and physical capabilities, synthesised within transitions that update the agent's state in response to reasoning, sensing and acting. Transitions are orchestrated by cycle theories that specify the order in which transitions are executed while taking into account the dynamic context and agent preferences, as well as selection operators for providing inputs to transitions.

Marek Sergot's technical contributions range over different subjects. He has developed a series of novel ideas and formal methods bridging different research domains, such as artificial intelligence, computational logic, philosophical logic, legal theory, artificial intelligence and law, multi-agent systems and bioinformatics. By combining his background in logic and computing with his interest in the law, deontic logic, action, and related areas, and applying to all his capacity to understand the subtleties of social interactions and normative reasoning, Marek has been able to open new directions of research, and has been a reference, an inspiration, and a model for many researchers in the many fields in which he has worked.

We study how the problem of temporal projection can be formalized in terms of argumentation. In particular, we extend earlier work of translating the language $\mathcal{E}$ for Reasoning about Actions and Change into a Logic Programming argumentation framework, by introducing new types of arguments for (i) backward persistence and (ii) persistence from observations. The paper discusses how this extended argumentation formulation is close to the original Event Calculus proposed by Kowalski and Sergot in 1986.

Many works have identified the potential benefits of using argumentation to address a large variety of multiagent problems. In this paper we take this idea one step further and develop the concept of a fully integrated argumentation-based agent architecture that allows us to develop agents that are coherently designed on an underlying argumentation based foundation. Under this architecture, an agent is composed of a collection of modules each of which is equipped with a local argumentation theory. Similarly, the intra-agent control of the agent is governed by local argumentation theories that are sensitive to the current situation of the agent through dynamically enabled feasibility arguments.

We propose a framework that brings together two major forms of default reasoning in Artificial Intelligence: default property classification in static domains, and default property persistence in temporal domains. Emphasis in this work is placed on the qualification problem , central when dealing with default reasoning, and in any attempt to integrate different forms of such reasoning. Our framework can be viewed as offering a semantics to two natural problems: (i) that of employing default static knowledge in a temporal setting, and (ii) the dual one of temporally projecting and dynamically updating default static knowledge. The proposed integration is introduced through a series of example domains, and is then formalized through argumentation. The semantics follows a pragmatic approach. At each time-point, an agent predicts the next state of affairs. As long as this is consistent with the available observations, the agent continues to reason forward. In case some of the observations cannot be explained without appealing to some exogenous reason, the agent revisits and revises its past assumptions. We conclude with some formal results, including an algorithm for computing complete admissible argument sets, and a proof of elaboration tolerance , in the sense that additional knowledge can be gracefully accommodated in any domain.

We argue that Abduction is useful for the problem of Knowledge Assimilation and use it to study the problems of Belief Revision and Truth Maintenance (TM) associated with the task of assimilating a series of observations Q1,...Qn. The close connection between Abduction and TM shown in [Reiter&de Kleer87] is explored further to provide a non-monotonic extension of ATMS that incorporates a dependency-directed backtracking mechanism thus combining the capabilities of ATMS and Doyle's TMS. This TM system has a well-defined semantics inherited from the semantics for abduction which is defined through a generalization of stable models based on autoepistemic logic.

This paper presents a novel approach to modelling embedded agent dialogues. It proposes a specific structure for the supporting information accompanying the arguments that agents exchange during a dialogue, it defines formally how this information relates to the agent theory, and assigns to it semantics that is associated to each of the atomic dialogue types of the Walton-Krabbe typology. This allows the formal definition of necessary and sufficient initiation and acceptance conditions of licit dialectical shifts that are necessary for the modelling of embedded agent dialogues.

We describe $\mathcal{M}$odular-e ($\mathcal{ME}$), a specialized, model-theoretic logic for narrative reasoning about actions, able to represent non-deterministic domains involving concurrency, static laws (constraints) and indirect effects (ramifications). We give formal results which characterize $\mathcal{ME}$'s high degree of modularity and elaboration tolerance, and show how these properties help to separate out, and provide a principled solutions to, the endogenous and exogenous qualification problems. We also show how a notion of (micro) processes can be used to facilitate reasoning at the dual levels of temporal granularity necessary for narrative-based domains involving “instantaneous” series of indirect and knock-on effects.

In this work, we extend the architecture of agents (and robots) based upon fixed, one-size-fits-all cycles of operation, by providing a framework of declarative specification of agent control. Control is given in terms of cycle theories, which define in a declarative way the possible alternative behaviours of agents, depending on the particular circumstances of the (perceived) external environment in which they are situated, on the internal state of the agents at the time of operation, and on the agents' behavioural profile. This form of control is adopted by the KGP model of agency and has been successfully implemented in the PROSOCS platform. We also show how, via cycle theories, we can formally verify properties of agents' behaviour, focusing on the concrete property of agents' interruptibility. Finally, we give some examples to show how different cycle theories give rise to different, heterogeneous agents' behaviours.

This paper proposes a model for intelligent navigation through multi-contextual information that could form electronic roads in the information society. This paper aims to address the problem of electronic information roads, define their notion and the technical form they can take as well as present the tools developed for implementing such a system. The main objective of the proposed model is to give the traveler the capability of exploring the information space in a natural way where the information offered will remain continuously interesting. The system offers links to information in a dynamic and adaptive way. This is achieved by employing intelligent navigation techniques, which combine user profiling and meta-data. Electronic roads emphasize the presentation of multi-contextual information, i.e., information that is semantically related but of different nature at different locations and time. An electronic road is the user’s navigation path through a series of information units. Information units are the building blocks of the available cultural information content.

Abduction in Logic Programming started in the late 80s, early 90s, in an attempt to extend logic programming into a framework suitable for a variety of problems in Artificial Intelligence and other areas of Computer Science. This paper aims to chart out the main developments of the field over the last ten years and to take a critical view of these developments from several perspectives: logical, epistemological, computational and suitability to application. The paper attempts to expose some of the challenges and prospects for the further development of the field. © 2002 Springer-Verlag Berlin Heidelberg. Book subtitle: ESSAYS IN HONOUR OF ROBERT A KOWALSKI ispartof: Computational Logic: Logic Programming and Beyond Essays in Honour of Robert A. Kowalski pages:402-436 ispartof: Lecture Notes in Artificial Intelligence vol:2407 pages:402-436 status: published

The A-system [4] is a new system for performing abductive reasoning within the framework of Abductive Logic Programming (ALP). The principles behind the system are founded by work on two earlier systems ACLP [2],[3] and SLDNFA(C) [1],[6]. The basic inference mechanism of the system combines abductive logicp rogramming and constraint logicp rogramming. In its computation it reduces the high level specification of the problem and goal at hand to a lower level constraint store. This constraint store is managed by an efficient constraint solver returning information to the abductive reduction process in order to help this in its search for a solution.

The language Ɛ for reasoning about actions and change can be translated into an argumentation framework. In this paper, we extend this translation of the basic language and showh ow it can, together with methods from abduction, form the basis for a principled implementation of Ɛ. The extension we have considered concerns the addition of new type of sentences in the language as well as allowing theories where the narrative of events given is incomplete. A system, called Ɛ-RES, is developed within the argumentation framework of Logic Programming without Negation as Failure (LPwNF). This can support directly a variety of modes of common sense reasoning such as: default persistence in credulous or sceptical form, assimilation of observations and their diagnosis possibly under incomplete information, as well as combinations of these. To improve the efficiency of the system we have considered the integration of a SAT solver within the LPwNF computation, to carry out the of validating the time universal constraints imposed by ramification statements.

The outline of this chapter is as follows. In Section 1.2 we discuss the philosophical and logical origins of abduction and induction. In Section 1.3 we analyse previous work on abduction and induction in the context of logic programming\indexlogic programming and artificial intelligence, and attempt a (partial) synthesis of this work. Section 1.4 considers the integration of abduction and induction in artificial intelligence, and Section 1.5 concludes.

We investigate how abduction and induction can be integrated into a common learning framework. In particular, we consider an extension of Inductive Logic Programming (ILP) for the case in which both the background and the target theories are abductive logic programs and where an abductive notion of entailment is used as the basic coverage relation for learning. This extended learning framework has been called Abductive Concept Learning (ACL). In this framework, it is possible to learn with incomplete background information about the training examples by exploiting the hypothetical reasoning of abduction. We also study how the ACL framework can be used as a basis for multiple predicate learning. An algorithm for ACL is developed by suitably extending the top-down ILP method: the deductive proof procedure of Logic Programming is replaced by an abductive proof procedure for Abductive Logic Programming. This algorithm also incorporates a phase for learning integrity constraints by suitably employing a system that learns from interpretations like ICL. The framework of ACL thus integrates the two ILP settings of explanatory (predictive) learning and confirmatory (descriptive) learning. The above algorithm has been implemented into a system also called ACL Several experiments have been performed that show the effectiveness of the ACL framework in learning from incomplete data and its appropriate use for multiple predicate learning.

We show how a class of domains written in the Language E, a high level language for reasoning about actions, narratives and change, can be translated into the argumentation framework of Logic Programming without Negation as Failure (LPwNF). This translation enables us 1) to understand default persistence as captured by various temporal reasoning frameworks in a simple and natural way, by assigning higher priority to information about effects of later actions over effects of earlier actions; 2) to develop an argumentation-based computational model for this type of reasoning in logic programming goal-driven style.

We present an approach for solving some of the problems of top-down Inductive Logic Programming systems when learning multiple predicates. The approach is based on an algorithm for learning abductive logic programs. Abduction is used to generate additional information that is useful for solving the problem of global inconsistency when learning multiple predicates.

The objective of this study is to investigate an approach for dynamic construction of Electronic Roads. We envision Electronic Roads spanning a virtual multidimensional space, a distributed digital information repository, comprising primarily of video data of cultural heritage. A visitor (user) will be able to travel in this multidimensional space along different, semantically related historical, geographical, economic or cultural paths.

We investigate how abduction and induction can be integrated into a common learning framework through the notion of Abductive Concept Learning (ACL). ACL is an extension of Inductive Logic Programming (ILP) to the case in which both the background and the target theory are abductive logic programs and where an abductive notion of entailment is used as the coverage relation. In this framework, it is then possible to learn with incomplete information about the examples by exploiting the hypothetical reasoning of abduction. The paper presents the basic framework of ACL with its main characteristics. An algorithm for an intermediate version of ACL is developed by suitably extending the top-down ILP method and integrating this with an abductive proof procedure for Abductive Logic Programming (ALP). A prototype system has been developed and applied to learning problems with incomplete information.

In this paper we present a framework for learning non-monotonic logic programs. The method is parametric on a classical learning algorithm whose generated rules are to be understood as default rules. This means that these rules must be tolerant to the negative information by allowing for the possibility of exceptions. The same classical algorithm is then used to learn recursively these exceptions. We prove that the non-monotonic learning algorithm that realizes these ideas converges asymptotically to the concept to be learned. We also discuss various general issues concerning the problem of learning nonmonotonic theories in the proposed framework.

In this paper, we study how a logical form of scientific modelling that integrates together abduction and induction can be used to understand the functional class of unknown enzymes or inhibitors. We show how we can model, within Abductive Logic Programming (ALP), inhibition in metabolic pathways and use abduction to generate facts about inhibition of enzymes by a particular toxin (e.g. Hydrazine) given the underlying metabolic pathway and observations about the concentration of metabolites. These ground facts, together with biochemical background information, can then be generalised by ILP to generate rules about the inhibition by Hydrazine thus enriching further our model. In particular, using Progol 5.0 where the processes of abduction and inductive generalization are integrated enables us to learn such general rules. Experimental results on modelling in this way the effect of Hydrazine in a real metabolic pathway are presented.