Area Exam

Ontology Databases

In this talk, I will explain how to align basic features of an ontology with the capabilities of relational databases. Besides using databases to store large volumes of data instances for an ontology, we can also offload some of the reasoning tasks typically handled by the knowledge base's inference engine. How far can we push this relationship between ontologies and databases remains an interesting question.

Peer-to-Peer Content Delivery

In this survey, we present a comprehensive taxonomy of content delivery mechanisms over the Internet that incorporates two key application requirements, namely the type of content (i.e. elastic and streaming content), and communication model (i.e one/many to one/many). Our taxonomy distinguishes between two approaches to implement a content delivery mechanism, which are placing the required machinery in the core or at the edge of the network.

Ontology-driven multidisciplinary problem solving environments

An important class of scientific problems are called {\it multidisciplinary} because they require information from different disciplines to approach them. Current problem solving environments (PSE) and workflow management systems are mature enough for single domain-oriented problem-solving. Addressing scientific problems that are multidisciplinary in nature are more challenging. The present approach to multidisciplinary scientific problems is through collaboratories and files sharing to support conventional human interactions between domain experts.

Internet Worm Detection

The occurrence of self-propagating network worms has decreased in recent years, but the threat that they pose to the Internet has not. Recent advances in polymorphism techniques, allergy attacks, and scanning strategies have rendered many existing worm detection mechanisms obsolete. In this talk we will examine Internet worms and their history, survey the current state-of-the-art in worm detection, and evaluate the strengths and weaknesses of each detection system. We will then discuss future research directions and important open questions in the field.

Identity Assurance on the Internet

Identity assurance is crucial to the secure operation of the Internet. Unfortunately, attackers today can utilize spoofed identities at multiple layers of the network, including the data link layer, network layer, and application layer. While identity assurance is important at all layers of the network, ensuring the identity at the network layer is critical to all communication on the Internet, since all Internet traffic uses the Internet Protocol (IP).

Ontology-Based Information Extraction on PubMed abstracts using the OMIT ontology to discover inconsistencies

Information Extraction (IE) aims to retrieve certain types of information from natural language text by processing them automatically. Ontology-Based Information Extraction (OBIE) has recently emerged as a subfield of Information Extraction. Here, ontologies - which provide formal and explicit specifications of conceptualizations - play a crucial role in the information extraction process. Because of the use of ontologies, this field is related to Knowledge Representation and has the potential to assist the development of the Semantic Web.

P2P Overlay, Internet Underlay and Their Mutual Impact

In P2P applications, participating peers form an overlay network through which they exchange data. Construction of the P2P overlays is often without any knowledge of the underlying network. The overlay impacts the underlay by the volume and pattern of the imposed traffic and the underlay may limit, block or otherwise affect the traffic flows associated with the overlay. The mutual impact between the overlay and the underlay is the main focus of this survey.

Financial Forecasting with Multiple Learning Models and Data Sources

Machine Learning has been applied to various, wide-ranging domains for decades, such as, bioinformatics, finance and image processing. One of the major open questions is how to learn knowledge from multiple data sources effectively and efficiently. The motivation of learning from multiple data sources is to obtain more accurate knowledge while understanding the relationships among data sources. Boosting, ensembling and multiple kernel learning were developed to address the above challenge in different scenarios.

Computational Models for Scientific Workflow

Scientific workflow languages and their supporting problem-solving environments enable researchers to effectively describe, deploy, monitor and control experiments and scientific procedures. Although several computational models may be used to describe workflow programs and their execution, the most common and effective have proven to be dataflow models and process calculi. We describe the important characteristics of scientific workflow and consider ways to specify and model the coordination and orchestration of the tasks within such workflows.

Semantic Data Mining

Incorporating domain knowledge is one of the most challenging problems in data mining. Such information is crucial in virtually all aspects of the data mining process including appropriate feature and model selection, reduction of the search space of hypotheses, representation of the output, and improvement of performance.


Subscribe to RSS - Area Exam