Abstract: This paper is about reducing influence diagram (ID) evaluation into Bayesian network (BN) inference problems that are as easy to solve as possible. Such reduction is interesting because it enables one to readily use one's favorite BN inference algorithm to efficiently evaluate IDs. Two such reduction methods have been proposed previously (Cooper 1988; Shachter and Peot 1992). This paper proposes a new method. The BN inference problems induced by the new method are much easier to solve than those induced by the two previous methods.

Abstract: A rational agent (artificial or otherwise) residing in a complex changing environment must gather information perceptually, update that information as the world changes, and combine that information with causal information to reason about the changing world. Using the system of defeasible reasoning that is incorporated into the OSCAR architecture for rational agents, a set of reason-schemas is proposed for enabling an agent to perform some of the requisite reasoning. Along the way, solutions are proposed for the Frame Problem, the Qualification Problem, and the Ramification Problem. The principles and reasoning described have all been implemented in OSCAR.

Abstract: We investigate a formal representation of time units, calendars, and time unit instances as restricted temporal entities for reasoning about repeated events. We generalize Allen's interval relations to a class level, and based on interval classes we define time units. We examine characteristics of time units, and provide a categorization of the hierarchical relations among them. Hence we define an abstract hierarchical unit structure (a calendar structure) that expresses specific relations and properties among the units that compose it. Specific objects in the time line are represented based on this formalism, including nonconvex intervals corresponding to repeated events. A goal of this research is to be able to represent and reason efficiently about repetition in time.

Abstract: The goal of our work is to develop theoretical foundations for the representation of knowledge in domains in which properties may vary continuously. One achievement of our research is that it extends the applicability of current research on theories of action. Furthermore, we are able to apply known approaches to the frame and ramification problems, developed for discretely changing worlds, to domains in which the world changes continuously. Our approach is based on the discrete situation calculus and on a monotonic solution to the frame problem. In order to address the combined frame and ramification problems, we extend Lin and Reiter's work. We use Pinto and Reiter's extension to the situation calculus to represent occurrences. We extend this work further to allow for reasoning by default. For example, if we know that a ball is falling and we do not have any reason to believe that an action would interfere with the ball's motion, then we assume that the ball will hit the ground. Finally, we extend the language of the situation calculus to allow for properties that change within situations. We also show that our proposed situation calculus inherits the solutions to the frame and ramification problems.

Abstract: Halpern has recently claimed a counterexample to Cox's Theorem, a well-known existence result for subjective probability distributions, but stated that the counterexample can be defeated by a specific assumption. Cox made this assumption, and so escapes the counterexample. Although Halpern questioned whether the assumption is reasonable for finite sets of sentences, it supports features that distinguish Cox's work from other, more restrictive motivations of probabilism. Paris has recently offered a new proof of Cox's Theorem whose correctness is satisfactory to Halpern, one that depends on a premise consistent with Cox's later work. As with any deductive argument, denial of a premise licenses denial of the conclusion, but Cox's conclusion does follow from premises plainly acceptable to him.

Abstract: Taxonomies (partially ordered sets and lattices) are important in many areas of computing science, particularly object-oriented languages, machine learning, and knowledge representation Taxonomic encoding strives to enhance the efficiency of taxonomic representation and use, which becomes increasingly important as the size of taxonomies grows In this paper, we describe a formal structure, called a spanning set, in which taxonomic encoding techniques can be characterized Any taxonomic encoding scheme implements a mapping from the original ordered set into a structure, such as the lattice of bit-vectors or logical terms, in which operations can be performed efficiently We analyze the fundamental properties any such mapping must satisfy in order to preserve subsumption, joins, or meets Spanning sets are an abstract framework within which we portray and compare existing encoding techniques, and provide a context in which new encoding problems can be analyzed, leading to existing, related algorithms or, using other results we develop in this paper, guiding the development of new algorithms We also explore the limits of minimal-sized encodings, proving a lower bound for simple forms of encoding and showing that, in general, finding minimal-sized encodings is NP-hard This paper can thus be viewed as both a synthesis of current research in taxonomic encoding and a repository of new results and directions for encoding as viewed from the perspective of spanning sets

Abstract: The focus of this paper is on an action representation formalism that encodes bothlinguistic andplanning knowledge about actions, and that supports the interpretation of complex Natural Language instructions and, in particular, of instructions containing Purpose Clauses. The representation uses linguistically motivated primitives, derived from Jackendoff's work on Conceptual Semantics, and is embedded in the description logic based system CLASSIC. I first motivate the characteristics of the formalism as needed to understand Natural Language instructions. I then describe the formalism itself, and I argue that the integration of a linguistically motivated lexical semantics formalism and of a description logic based system is beneficial to both. Finally, I show how the formalism is exploited by the algorithm that interprets Purpose Clauses. The output of the algorithm is used in theAnimation from NL project, that has as its goal the automatic creation of animated task simulations.

Abstract: Shapes such as triangles or rectangles can be defined in terms of geometric properties invariant under a group of transformations. Complex shapes can be described by logic formulas with simpler shapes as the atoms. A standard technique for computing invariant properties of simple shapes is the method of moment invariants, known since the early 1960s. We generalize this technique to shapes described by arbitrary monotone formulas (formulas in propositional logic without negation). Our technique produces a reduced Grobner basisfor approximate shape descriptions. We show how to use this representation to solve decision problems related to shapes. Examples include determining if a figure has a particular shape, if one description of a shape is more general than another, and whether a specific geometric property is really necessary for specifying a shape. Unlike geometry theorem proving, our approach does not require the shapes to be explicitly defined. Instead, logic formulas combined with measurements performed on actual shape instances are used to compute well-characterized least squares approximations to the shapes. Our results provide a proof that decision problems stated in terms of these approximations can be solved in a finite number of steps.

Abstract: Locational reasoning plays an important role in many applications of AI problem-solving systems, yet has remained a relatively unexplored area of research. This paper addresses both theoretical and practical issues relevant to reasoning about locations. We define several theories of location designed for use in various settings, along with a sound and complete belief revision calculus for each that maintains a STRIPS-style database of locational facts. Techniques for the efficient operationalization of the belief revision rules in planning frameworks are presented. These techniques were developed during application of the location theories to several large-scale planning tasks within the Sipe planning framework.

Abstract: This paper presents the development of neural controllers for the swimming and the trotting of a salamander-like animat. Using a genetic algorithm (GA), we extend a connectionist model of the central pattern generator (CPG) controlling the swimming of a lamprey 1] to control the locomotion of a 2D mechanical simulation of a salamander. We study in particular what kind of neural connectiv-ity can produce the traveling undulation of the trunk during swimming and the standing S-wave undulation during trotting. Using a GA, we evolve neural controllers which are made of two segmental oscillators controlling the fore-and hindlimb muscles, which project to a lamprey-like CPG for the trunk muscles. CPGs are successfully evolved which exhibit the swimming gait when external excitatory drive is applied only to the trunk CPG, and the trotting gait when external drive is applied to both the limb and the trunk CPGs. For both types of gaits, the speed of locomotion can be varied with the amplitude of the excitatory drive, and the direction of motion can be changed when asymmetrical drive is applied.

Abstract: The goal of this paper is to motivate and define yet another sorted logic, called SND. All the previous sorted logics that can be found in the Artificial Intelligence literature have been designed to be used in (completely) automated deduction. SND has been designed to be used in interactive theorem proving. Because of this shift of focus, SND has been designed to satisfy three innovative design requirements: it is defined on top of a natural deduction calculus, and in a way to be a definitional extension of such calculus; and it is implemented on top of its implementation. In turn, because of this fact, SND has various innovative technical properties; among them: it allows us to deal with free variables, it has no notion of well-sortedness and of well-sortedness being a prerequisite of well-formedness, its implementation is such that, in the default mode, the system behaves exactly as with the original unsorted calculus.

Abstract: The paradox of the preface and the lottery paradox are paradoxes of practical certainty sharing certain features. The paradox of the lottery argues that rational agents are at once practically certain that each ticket in a lottery will lose but also practically certain some ticket will win. The paradox of the preface argues that rational agents are at once practically certain that all facts in a written volume are true, yet are also practically certain that some fact is wrong. A difference between real lotteries and prefaces is that a winning lottery ticket is generally an intended feature of the lottery, whereas incorrect facts are generally unintended. Despite these similarities, Pollock gives a novel argument suggesting that the preface paradox warrants qualitatively different treatment from the lottery, using as a rationale the differences between real lotteries and prefaces. This draws a clear line between the work of Pollock and the work of Kyburg, both of whom have had a prominent influence in recent thinking on nonmonotonic reasoning in AI. This note shows there are real lotteries with the formal structure of the preface paradox and possibly prefaces with the formal structure of lotteries. The surprising conclusion is that within Pollock's framework, the treatment of any problem with a formal structure resembling the lottery (or the preface) depends on the process by which winning tickets (or publishing errors) are generated. The rationales given by Pollock seem to be unrelated to the actual mechanisms implemented.

Abstract: In a real-time task an action must be executed given limited computation. One approach to limited computation is to search a tree of possible action sequences to a fixed depth and then execute an action with the lowest associated backed-up cost. The standard algorithm for such a search is Depth-First Branch-and-Bound (DFBB), also known in the Artificial Intelligence literature as Minimin with Alpha Pruning. This article shows that a depth-bounded extension of a popular iterative algorithm called IDA has a surprisingly large range of search trees on which it outperforms DFBB—something previous analytical results do not predict. We prove that the extended algorithm, which we call DIDA, is correct, is guaranteed to terminate, and is asymptotically (i.e., on its last iteration) as efficient as DFBB—assuming a consistent heuristic is used. We also prove that both algorithms are guaranteed not to decrease their accuracy with a deeper search, assuming a consistent heuristic. Because accuracy is generally correlated with decision quality, the time saved by visiting fewer states translates to deeper searches which translates to better decisions. Results from random search trees show that DIDA is most efficient when the path cost + leaf node heuristic value is distributed with low variance; as branching factor increases, the range for which DIDA is more efficient also increases. Results with Eight, Fifteen, Twenty-four, and Ninety-nine Puzzle implementations of both algorithms—all domains with low variance of path cost + leaf node heuristic value—show that DIDA significantly outperforms DFBB.

Abstract: We present in this paper a cortically-inspired model that learns to exploit regularities of sequences of perceptions and actions, with regard to motivations. Sequences are learned from continuous multimodal information streams, on the basis of a competition mechanism. This approach enables generalisations from perceptive regularities and also ensures the capability to apply results of previous learning to help further learning of new tasks.