This thesis demonstrates how the power of symbolic processing can be exploited in the learning of low level control functions. It proposes a novel hybrid architecture with a tight coupling between a variant of symbolic planning and reinforcement learning. This architecture combines the strengths of the function approximation of subsymbolic learning with the more abstract compositional nature of symbolic learning. The former is able to represent mappings of world states to actions in an accurate way. The latter allows a more rapid solution to problems by exploiting structure within the domain. A control function is learnt over time through interaction with the world. Symbols are attached to features in the functions. The symbolic attachments act as anchor points used to transform the function of a previously learnt task to that of a new task. The solution of more complex tasks is achieved through composing simpler functions, using the symbolic attachments to determine the composition. The result is used as the initial control function of the new task and then modified through further learning. This is shown to produce a significant speed up over basic reinforcement learning.

In many domains, characterizations of a given attribute are imprecise, uncertain and incomplete in the available learning examples. The definitions of classes may be vague. Learning systems are frequently forced to deal with such uncertainty. Traditional learning systems are designed to work in the domains where imprecision and uncertainty in the data are absent. Those learning systems are limited because of their impossibility to cope with uncertainty--a typical feature of real-world data. In this thesis, we developed a fuzzy learning system which combines inductive learning with a fuzzy approach to solve problems arising in learning tasks in the domains affected by uncertainty and vagueness. Based on Fuzzy Logic, rather than pure First Order Logic used in FOIL, this system extends FOIL with learning fuzzy logic relation from both imprecise examples and background knowledge represented by Fuzzy Prolog. The classification into the positive and negative examples is allowed to be a degree (of positiveness or negativeness) between 0 and 1. The values of a given attribute in examples need not to be the same type. Symbolic and continuous data can exist in the same attribute, allowing for fuzzy unification (inexact matching). An inductive learning problem is formulated as to find a fuzzy logic relation with a degree of truth, in which a fuzzy gain calculation method is used to guide heuristic search. The Fuzzy FOIL's ability of learning the required fuzzy logic relations and dealing with vague data enhances FOIL's usefulness.

This thesis addresses the problem of learning concept descriptions that are interpretable, or explainable. Explainability is understood as the ability to justify the learned concept in terms of the existing background knowledge. The starting point for the work was an existing system that would induce only fully explainable rules. The system performed well when the model used during induction was complete and correct. In practice, however, models are likely to be imperfect, i.e. incomplete and incorrect. We report here a new approach that achieves explainability with imperfect models. The basis of the system is the standard inductive search driven by an accuracy-oriented heuristic, biased towards rule explainability. The bias is abandoned when there is heuristic evidence that a significant loss of accuracy results from constraining the search to explainable rules only. The users can express their relative preference for accuracy vs. explainability. Experiments with the system indicate that, even with a partially incomplete and/or incorrect model, insisting on explainability results in only a small loss of accuracy. We also show how the new approach described can repair a faulty model using evidence derived from data during induction.

As more sophisticated systems are being developed, powerful approaches for modeling their behavior and test their reliability are necessary. The research work in this thesis takes on the problem of building a Graphical Programming Environment that permits to create models of DESs in a timed temporal logic framework and simulate the DES models in real-time using an object oriented environment through the interconnection of visual symbols. A temporal logic framework is developed to write the formal models of the temporal references of DESs. This approach is enhanced by the inclusion of a global clock variable to add real-time properties to the formal specifications of real-time DESs. The interactive visual environment allows the programmer to activate graphical symbols by means of menu selections. The graphical symbols are grouped into classes which are eventually properly interconnected, parsed and mapped into source code written in the timed temporal logic language. A knowledge-based system is composed of knowledge databases (database of facts and database of rules), These databases, representing the system behavior, can be created using this tool, for which a reasoning mechanism is required. An inference engine is designed to interpret these knowledge databases. An OO programming language is used, Objective-C. It is used throughout the design, however, when using the tool, the user does not notice the underlying programming language, in other words, the programming language is transparent to the user. The Graphical Programming Environment designed in this thesis can be used to build the specifications of real-time DESs. Different knowledge databases have been created using this interactive tool for three examples to verify their behaviors, such examples are: The ABP communication protocol, the packet-switched communication protocol, and the telephone system.

Over the past few years, the subject of Intelligent Network (IN) has captured the interest of the telecommunications community. The objective of IN is to allow the introduction of new capabilities in the telecommunications network and to facilitate and accelerate in a cost-effective manner, service implementation and provisioning, in a multivendor environment. However, this objective confronts a major obstacle known as the feature interaction problem. The feature interaction problem occurs when a feature is prevented from performing its functionalities in the presence of other features. In the first part of the thesis, we describe a LOTOS model for structuring the Functional Entities (FEs) that are defined in the Distributed Functional Plane (DFP) of the CS1 IN Conceptual Model (INCM), and that are involved in the establishment of a call/connection and invocation and processing of services. The specification of IN services is achieved using Service Independent building Blocks (SIBs). It is designed in a way that independent specification and rapid introduction of services is provided. In the second part of the thesis, a method for detecting feature interactions between services is developed. The method is limited to the detection of interactions caused by violation of features properties. It is based on formalization of feature's properties, derivation of goals satisfying the negation of these properties and use of Goal Oriented Execution to detect traces satisfying these goals. A trace satisfying a goal shows that an interaction exists between the specified features by describing a scenario violating one of the properties of the introduced features. It is concluded that LOTOS is useful as a Formal Description Technique (FDT) in the Service Creation Environment (SCE). The developed specification can be used for adding specifications of new services, and for detecting interactions caused by violation of properties, if there are any.

Inductive Logic Programming (ILP) is one of the new and fast growing sub-fields of artificial intelligence. Given a specification language, the goal is to induce a logic program from examples of how the program should work (and also of how it should not work). One main difficulty of ILP lies in learning recursively defined predicates. Today's systems strongly rely on a set of supporting predicates known as the background knowledge that helps define the recursive clause. The dependence on background knowledge has its drawbacks in that it is assumed that the user knows in advance what sort of predicates are required by the target definition. Predicate invention, a research topic that has received a lot of attention lately, can remedy the situation by extending the specification language with new concepts, which appear neither in the examples nor in the background knowledge, and finding a definition for them. A serious concern is that no examples of the invented predicate are explicitly given but rather of the target predicate, so learning has to be done in the absence or scarcity of examples. This research is concerned with the problem of learning recursive definitions based on inverting clausal implication from a small data set. The aim is both to derive an autonomous learning method that can invent the recursive predicates it needs, and to implement it in an efficient manner. Experiments show that the system is capable of finding a correct definition of many relations by inventing the necessary predicates, but does not perform very well on random examples. A comparison between several similar systems that learn recursive definitions of a single predicate is shown. We also show the need for system-generated negative examples and discuss several pitfalls of predicate invention and the absence/scarcity of examples.

The industry has widely accepted the concept of design patterns to promote quality design reuse in the recent years. However, there are several problems preventing design patterns being used efficiently and effectively. The design pattern recognition system, DEPARS, discussed in this dissertation relieves these problems and promotes design pattern reuse. DEPARS recognizes patterns in object models by matching to templates in the knowledge base. DEPARS arranges the templates in the knowledge base in a hierarchy such that templates close to the root of the hierarchy are the bases of the ones below. The hierarchy reduces DEPARS's matching effort because it narrows the search area. DEPARS provides information about the recognized patterns to designers. This information helps designers to apply appropriate patterns in designs. DEPARS has pattern mining capability. DEPARS recognizes new patterns that may be reusable in the future from existing designs. In addition, DEPARS also facilitates designers verifying the recurrence of proto-patterns by storing the proto-patterns in the knowledge base. Once the proto-patterns are in the knowledge base, DEPARS can recognize them in future designs and hence shows the recurrence of the proto-patterns. The dissertation presents the design and operation of DEPARS. The dissertation also reports and discusses the evaluation results of DEPARS. The evaluation shows promising results indicating that DEPARS is adequate for practical use.

The Intelligent Network (IN) is a conceptual model for a service development technology to create telecommunication services. In its current form, IN is limited to service creation in isolated networks and cannot support co-operative service development between two or more networks. Rapid development in networking paradigms and standards has led to an urgent need of finding solutions to the problem of interworking heterogeneous networks. Differing abstraction levels make meaningful exchange of information difficult, and IN has not been able to meet this requirement. The Reference Model for Open Distributed Processing (RM-ODP) is a distributed object based architecture which provides a high level framework for distributed systems. The emphasis is to develop a set of re-usable functional abstractions that can be recombined in various configurations to develop required applications. This work uses RM-ODP framework to supplement deficiencies evident in IN. Two specific aspects are examined and developed. The first is service portability through service profile modeling. A model for service development in a mobile environment, and related concepts of service profile modeling and transfer are developed. The second, IN domain interworking in the ODP framework. A ODP framework for the modeling of this service profile and its migration as the user moves to different domains is proposed. Our approach allows dynamically configured interworking of domains.

This thesis is concerned with sequential decision making by multiple agents, whether they are acting cooperatively to maximize team reward or selfishly trying to maximize their individual rewards. The practical intractability of this general problem led to efforts in identifying special cases that admit efficient computation, yet still represent a wide enough range of problems. In our work, we identify the class of problems with structured interactions, where actions of one agent can have non-local effects on the transitions and/or rewards of another agent. We addressed the following research questions: (1) How can we compactly represent this class of problems? (2) How can we efficiently calculate agent policies that maximize team reward (for cooperative agents) or achieve equilibrium (self-interested agents)? (3) How can we exploit structured interactions to make reasoning about communication offline tractable? For representing our class of problems, we developed a new decision-theoretic model, Event-Driven Interactions with Complex Rewards (EDI-CR), that explicitly represents structured interactions. EDI-CR is a compact yet general representation capable of capturing problems where the degree of coupling among agents ranges from complete independence to complete dependence. For calculating agent policies, we draw on several techniques from the field of mathematical optimization and adapt them to exploit the special structure in EDI-CR. We developed a Mixed Integer Linear Program formulation of EDI-CR with cooperative agents that results in programs much more compact and faster to solve than formulations ignoring structure. We also investigated the use of homotopy methods as an optimization technique, as well as formulation of self-interested EDI-CR as a system of non-linear equations. We looked at the issue of communication in both cooperative and self-interested settings. For the cooperative setting, we developed heuristics that assess the impact of potential communication points and add the ones with highest impact to the agents’ decision problems. Our heuristics successfully pick communication points that improve team reward while keeping problem size manageable. Also, by controlling the amount of communication introduced by a heuristic, our approach allows us to control the tradeoff between solution quality and problem size. For self-interested agents, we look at an example setting where communication is an integral part of problem solving, but where the self-interested agents have a reason to be reticent (e.g. privacy concerns). We formulate this problem as a game of incomplete information and present a general algorithm for calculating approximate equilibrium profile in this class of games.

abstract: Image super-resolution (SR) is a low-level image processing task, which has manyapplications such as medical imaging, satellite image processing, and video enhancement, etc. Given a low resolution image, it aims to reconstruct a high resolution image. The problem is ill-posed since there can be more than one high resolution image corresponding to the same low-resolution image. To address this problem, a number of machine learning-based approaches have been proposed. In this dissertation, I present my works on single image super-resolution (SISR) and accelerated magnetic resonance imaging (MRI) (a.k.a. super-resolution on MR images), followed by the investigation on transfer learning for accelerated MRI reconstruction. For the SISR, a dictionary-based approach and two reconstruction based approaches are presented. To be precise, a convex dictionary learning (CDL) algorithm is proposed by constraining the dictionary atoms to be formed by nonnegative linear combination of the training data, which is a natural, desired property. Also, two reconstruction-based single methods are presented, which make use of (i)the joint regularization, where a group-residual-based regularization (GRR) and a ridge-regression-based regularization (3R) are combined; (ii)the collaborative representation and non-local self-similarity. After that, two deep learning approaches are proposed, aiming at reconstructing high-quality images from accelerated MRI acquisition. Residual Dense Block (RDB) and feedback connection are introduced in the proposed models. In the last chapter, the feasibility of transfer learning for accelerated MRI reconstruction is discussed. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2020