<div> Graph representation learning and Graph Neural Network (GNNs) models provide flexible tools for modeling and representing relational data (graphs) in various application domains. Specifically, node embedding methods provide continuous representations for vertices that has proved to be quite useful for prediction tasks, and Graph Neural Networks (GNNs) have recently been used for semi-supervised node and graph classification tasks with great success. </div><div> </div><div> However, most node embedding methods for unsupervised tasks consider a simple, sparse graph, and are mostly optimized to encode aspects of the network structure (typically local connectivity) with random walks. And GNNs model dependencies among the attributes of nearby neighboring nodes rather than dependencies among observed node labels, which makes it not expressive enough for semi-supervised node classification tasks. </div><div> </div><div> This thesis will investigate methods to address these limitations, including: </div><div><br></div><div> (1) For heterogeneous graphs: Development of a method for dense(r), heterogeneous graphs that incorporates global statistics into the negative sampling procedure with applications in recommendation tasks;</div><div> (2) For capturing long-range role equivalence: Formalized notions of representation-based equivalence w.r.t regular/automorphic equivalence in a single graph or multiple graph samples, which is employed in a embedding-based models to capture long-range equivalence patterns that reflect topological roles; </div><div> (3) For collective classification: Since GNNs model dependencies among the attributes of nearby neighboring nodes rather than dependencies among observed node labels, we develop an add-on collective learning framework to GNNs that provably boosts their expressiveness for node classification tasks, beyond that of an {\em optimal} WL-GNN, utilizing self-supervised learning and Monte Carlo sampled embeddings to incorporate node labels during inductive learning for semi-supervised node classification tasks.</div>

<p>Multi-criteria decision making (MCDM) methods have evolved over the past several decades. In today’s world with rapidly growing industries, MCDM has proven to be significant in many application areas. In this study, a decision-making model is devised using reinforcement learning to carry out multi-criteria optimization problems. Learning automata algorithm is used to identify an optimal solution in the presence of single and multiple environments (criteria) using pareto optimality. The application of this model is also discussed, where the model provides an optimal solution to the food, energy, and water systems (FEWS) problem.</p>

No description available.

PDE models are a major tool used in quantitative modeling of biological and scientific phenomena. Their major shortcoming is the high computational complexity of solving each model. When scaling up to millions of simulations needed to find their optimal parameters we frequently have to wait days or weeks for results to come back. To cope with that we propose a neural network approach that can produce comparable results to a PDE model while being about 1000x faster. We quantitatively and qualitatively show the neural network metamodels are accurate and demonstrate their potential for multi-objective optimization in biology. We hope this approach can speed up scientific research and discovery in biology and beyond.<br>

When medical researchers test a new treatment procedure, they recruit patients with appropriate health problems and medical histories. An experiment with a new procedure is called a clinical trial. The selection of patients for clinical trials has traditionally been a labor-intensive task, which involves matching of medical records with a list of eligibility criteria. A recent project at the University of South Florida has been aimed at the automation of this task. The project has involved the development of an expert system that selects matching clinical trials for each patient. If a patient's data are not sufficient for choosing a trial, the system suggests additional medical tests. We report the work on the representation and entry of the related selection criteria and medical tests. We first explain the structureof the system's knowledge base, which describes clinical trials and criteria for selecting patients. We then present an interface that enables a clinician to add new trials and selection criteria without the help of a programmer. Experiments show that the addition of a new clinical trial takes ten to twenty minutes, and that novice users learn the full functionality of the interface in about an hour.

These are the proceedings of the International Workshop on Reactive Concepts in Knowledge Representation (ReactKnow 2014), which took place on August 19th, 2014 in Prague, co-located with the 21st European Conference on Artificial Intelligence (ECAI 2014).

Abstract dialectical frameworks (ADFs) are a formalism for representing knowledge about abstract arguments and various logical relationships between them. This work studies ADFs in detail. Firstly, we use the framework of approximation fixpoint theory to define various semantics that are known from related knowledge representation formalisms also for ADFs. We then analyse the computational complexity of a variety of reasoning problems related to ADFs. Afterwards, we also analyse the formal expressiveness in terms of realisable sets of interpretations and show how ADFs fare in comparison to other formalisms. Finally, we show how ADFs can be put to use in instantiated argumentation, where researchers try to assign meaning to sets of defeasible and strict rules. The main outcomes of our work show that in particular the sublanguage of bipolar ADFs are a useful knowledge representation formalism with meaningful representational capabilities and acceptable computational properties.

The field of artificial intelligence, research on knowledge representation and reasoning has originated a large variety of formats, languages, and formalisms. Over the decades many different tools emerged to use these underlying concepts. Each one has been designed with some specific application in mind and are even used nowadays, where the internet is seen as a service to be sufficient for the age of Industry 4.0 and the Internet of Things. In that vision of a connected world, with these many different formalisms and systems, a formal way to uniformly exchange information, such as knowledge and belief is imperative. That alone is not enough, because even more systems get integrated into the online world and nowadays we are confronted with a huge amount of continuously flowing data. Therefore a solution is needed to both, allowing the integration of information and dynamic reaction to the data which is provided in such continuous data-flow environments. This work aims to present a unique and novel pair of formalisms to tackle these two important needs by proposing an abstract and general solution. We introduce and discuss reactive Multi-Context Systems (rMCS), which allow one to utilise different knowledge representation formalisms, so-called contexts which are represented as an abstract logic framework, and exchange their beliefs through bridge rules with other contexts. These multiple contexts need to mutually agree on a common set of beliefs, an equilibrium of belief sets. While different Multi-Context Systems already exist, they are only solving this agreement problem once and are neither considering external data streams, nor are they reasoning continuously over time. rMCS will do this by adding means of reacting to input streams and allowing the bridge rules to reason with this new information. In addition we propose two different kind of bridge rules, declarative ones to find a mutual agreement and operational ones for adapting the current knowledge for future computations. The second framework is more abstract and allows computations to happen in an asynchronous way. These asynchronous Multi-Context Systems are aimed at modelling and describing communication between contexts, with different levels of self-management and centralised management of communication and computation. In this thesis rMCS will be analysed with respect to usability, consistency management, and computational complexity, while we will show how asynchronous Multi-Context Systems can be used to capture the asynchronous ideas and how to model an rMCS with it. Finally we will show how rMCSs are positioned in the current world of stream reasoning and that it can capture currently used technologies and therefore allows one to seamlessly connect different systems of these kinds with each other. Further on this also shows that rMCSs are expressive enough to simulate the mechanics used by these systems to compute the corresponding results on its own as an alternative to already existing ones. For asynchronous Multi-Context Systems, we will discuss how to use them and that they are a very versatile tool to describe communication and asynchronous computation.

We discuss a new concept of agent programs that combines logic programming with reasoning about actions. These agent logic programs are characterized by a clear separation between the specification of the agent’s strategic behavior and the underlying theory about the agent’s actions and their effects. This makes it a generic, declarative agent programming language, which can be combined with an action representation formalism of one’s choice. We present a declarative semantics for agent logic programs along with (two versions of) a sound and complete operational semantics, which combines the standard inference mechanisms for (constraint) logic programs with reasoning about actions.

<pre>Machine learning is widely used to extract meaningful information from video, images, audio, text, and other multimedia data. Through a hierarchical structure, modern neural networks coupled with backpropagation learn to extract information from large amounts of data and to perform specific tasks such as classification or regression. In this thesis, we explore various approaches to multimedia analytics with neural networks. We present several image synthesis and rendering techniques to generate new images for training neural networks. Furthermore, we present multiple neural network architectures and systems for commercial logo detection, 3D pose estimation and tracking, deepfakes detection, and manipulation detection in satellite images.<br></pre>