Knowledge representation techniques. A rough set approach. (Q2498271): Difference between revisions

Knowledge representation is one of the most important elements of Artificial Intelligence, representing the study of how knowledge about the world can be represented and what kinds of reasoning can be done with that knowledge. The book contains three parts and is founded on the concept of rough sets. The first part, entitled \textit{Introduction and Preliminaries}, has five chapters and presents: some basic notions regarding algebra of sets, metric spaces, computational complexity, propositional, predicate and fixpoint calculus, first-order, second-order and three-valued logic; upper and lower approximations as the basic concepts in rough set theory, decision systems and decision rules, dependencies between the attributes of two sets of attributes; the concepts of relational and deductive databases and various languages for representing rules and/or queries: predicate and fixpoint calculus, the languages Datalog, Datalog with negation and SHQL; nonmonotonic reasoning: default logic, circumscription. Part two, entitled \textit{From Relations to Knowledge Representation}, is composed of six chapters and studies: rough knowledge databases obtained by the combination of an extensional database, an intensional database and contextually closed queries: the architecture, the languages, the semantics, a method to compute the relation R obtained by applying the LCC (Local Contextual Closure) policy and preserving integrity constraints (in the general case and under a restriction on integrity constraints which allows one to compute explicit definitions of the new relations as first-order formulas and fixpoint formulas); approximation traducers which generate or define an approximate relation in terms of other relations from the input using various dependencies between the input and the output (defined in terms of upper and lower approximations), each dependency being a logical formula in a first-order logical language; the weakest sufficient condition that provides the best lower approximation and the strongest necessary condition that provides the best upper approximation on a formula \(A\) (expressed in the less expressive language) and some applications in communicating agents, approximation theory, abduction and generating successor state axioms in a robot domain; the method CAKE (Computed Aided Knowledge Engineering), which is viewed as an extension of well-known entity-relationship diagrams designed for representing relations in relational databases and is a means for constructing and visualizing the complex dependencies between granules: the language, diagrams, semantics and computation methods; the formalization of default logic using CAKE: two basic versions of default logic (rough default logic and rough logic with strong prerequisites) are represented by this method and it is shown that both versions can be extended with prioritization; a UAV (Unmanned Aerial Vehicles) scenario in order to identify and track all vehicles in a region and log the estimated velocities and positions of all small blue vehicles identified, for the duration of their stay in this region or until the UAV is low on fuel. Part three, entitled \textit{From Sensors to Relations}, contains four chapters and studies: granular computing: elementary granules, operations, extension of the CAKE method in order to provide diagrams for representing general information granules; tolerance and similarity measures which refer to conceptual spaces, tolerance functions and spaces, tolerance on elementary and arbitrary granules and applications to sensor models, decision rules, agents; a rough set methodology for supervised machine learning which includes: a general method for computing reducts in information and decision systems, constructing a high quality classifier, constructing minimal rules for decision systems, rough classifiers, applications of rough set techniques in order to extract association rules from information systems; UAV learning process which shows how, beginning with sensor data, one can generate decision rules, classifiers and rough relations that can be used for higher level reasoning in a robotic architecture. This book is recommended to researchers interested in studying and applying rough set theory in various domains.