Suk-Chung Yoon

Using domain knowledge in knowledge discovery

With the explosive growth of the size of databases, many knowledge discovery applications deal with large quantities of data. There is an urgent need to develop methodologies which will allow the applications to focus search to a potentially interesting and relevant portion of the data, which can reduce the computational complexity of the knowledge discovery process and improve the interestingness of discovered knowledge. Previous work on semantic query optimization, which is an approach to take advantage of domain knowledge for query optimization, has demonstrated that significant cost reduction can be achieved by reformulating a query into a less expensive yet equivalent query which produces the same answer as the original one. In this paper, we introduce a method to utilize three types of domain knowledge in reducing the cost of finding a potentially interesting and relevant portion of the data while improving the quality of discovered knowledge. In addition, we propose a method to select relevant domain knowledge without an exhaustive search of all domain knowledge. The contribution of this paper is that we lay out a general framework for using domain knowledge in the knowledge discovery process effectively by providing guidelines.

Semantic query processing in object-oriented databases using deductive approach

In this paper, we present a method to utilize semantic constraints that play an important role in search space reduction and termination of query evaluation in object-oriented databases. Our approach consists of three successive refinement steps: rule generation, semantic knowledge compilation, and semantic reformulation. In the rule generation step, we generate a set of deductive rules based on an object-oriented database schema or semantic knowledge about the domain of the database. In the semantic knowledge compilation step, we compile semantic knowledge together with object-oriented database schema to identify semantic knowledge that is potentially relevant(beneficial) to each class in the object-oriented database schema. We associate the fragments of valid and useful semantic knowledge, called residues with classes. After this step, semantic knowledge is grouped according to the classes that it references. During the semantic reformulation step, we receive a user’s query, select the set of relevant semantic knowledge in a query context, and transform the query with associated semantic knowledge into another form which is more efficiently processed. To our knowledge, there is no significant research that has been done about semantic query optimization in object-oriented databases using deductive approach. The unique contribution of this paper is that we extend semantic query optimization techniques developed for deductive databases to apply to object-oriented databases. Our approach attempts to minimize the number of operations that will be performed at run time.

Intelligent query answering in deductive and object-oriented databases

In the near future, we believe that we will need much more sophisticated answer-finding schemes in an object-oriented database in order to satisfy the needs of truly intelligent information system. In this paper, we introduce a method to apply the intensional query processing techniques of deductive databases to object-oriented databases. So, we can generate intensional answers to represent answer-set abstractly for a given query in object-oriented databases. Our approach consists of four steps: rule generation, pre-resolution, resolution, and post-resolution. In rule generation, we generate a set of deductive rules based on an object-oriented database schema. In pre-resolution, rule transformation is done to get unique intensional literals and extended term-restricted rules. In resolution, we identify rules that are potentially relevant to a query. In post-resolution, we find relevant resolvents as candidates for intensional answers among potentially relevant resolvents. We also use the notion of potentially relevant resolvents and relevant resolvents to avoid generating certain meaningless intensional answers.

An approach to intensional query answering at multiple abstraction levels using data mining approaches

Introduces a partially automated method for generating intensional answers at multiple abstraction levels for a query, which can help database users find more interesting and desired answers. Our approach consists of three phases: pre-processing, query execution and answer generation. In the pre-processing phase, we build a set of concept hierarchies constructed by generalization of the data stored in a database and a set of virtual hierarchies to provide a global view of the relationships among high-level concepts from multiple concept hierarchies. In the query execution phase, we receive a users query, process the query, collect an extensional answer and select a set of relevant attributes to be generalized in the extensional answer. In the answer generation phase, we find the general characteristics of those relevant attribute values at multiple abstraction levels with the concept hierarchies and the virtual hierarchies by using data mining methods. The main contribution of this paper is that we apply and extend data mining methods to generate intensional answers at multiple abstraction levels, which increases the relevance of the answers. In addition, we suggest strategies to avoid meaningless intensional answers, which substantially reduces the computational complexity of the intensional answer generation process.

Intensional query processing using data mining approaches

1 Introduction

Towards Conceptual Query Answering

With the increase in the volume and complexity of databases, we need much more sophisticated query-processing schemes in databases to satisfy the needs of truly intelligent user-machine interfaces required by new generation database applications. In this paper, we introduce a. partially witornaled method for conceptual query answering which is a inechanisin to answer queries specified with general and abstract terms rather than primitive data stored in databases. Conceptual query answering consists of two phases: preprocessing and execution. In the preprocessing phase, we discover useful and interesting abstract terms by building a set of concept hierarchies constructed by generalization of primitive data stored in a database into appropriate higher level concepts. Then, we construct an abs~racted database by generalizing and preprocessing primitive data in frequently referenced relations. Specifically, we find frequent conjuncts of the attributes which have meaningful correlations and replace the values of those attributes with the abstract terms defined in their concept hierarchies or results of aggregation functions on the values. In the execution phase, we receive a users conceptual query and process the query with the concept hierarchies and the abstracted database. The contribution of this paper is that we develop a framework for processing conceptual queries. In addtion, we suggest strategies to reduce the computational complexity of the conceptual query answer generation process.

Incremental versus non-incremental learning in adaptive common spatial patterns

Extracting reliable and discriminative features remains a critical challenge in the development of brain computer interface (BCI) techniques. Common spatial patterns (CSP) is frequently used for spatial filtering and feature extraction in electroencephalography (EEG)-based BCI. It performs a supervised and subject-specific learning of EEG data acquired in two different task conditions. Incremental learning has been used in CSP to adapt to a target subjects data by including classified data in training data and re-estimating spatial filters. In practical circumstances where no user feedback is instantly available to provide true class labels of target trials, misclassified EEG trials will be added to the training data of a wrong class, and potentially influence the training of spatial filters and feature extraction. In this study, incremental and non-incremental learning were investigated based on a recently developed adaptive CSP (ACSP) method using multi-subject EEG data. Their performances were compared in terms of intra- and inter-subject classification performances. Experimental results indicate that the non-incremental learning is a better option when true class labels of target data are not provided.

Developing Computer Forensics Minor -- Challenges and Opportunities: (Abstract Only)

This lightning talk will discuss our experience of developing and managing a new Computer Forensics Minor. The Computer Forensics minor is an interdisciplinary program that integrates criminal justice and computer science and combines both theoretical concepts and practical skills to prepare students for a career in computer forensics-related fields. Students will be prepared for a career in law enforcement or corporate security as a digital investigator and evidence examiner as well as pursue graduate education in the area of information security, digital forensics, or law. The lighting talk will describe the various stages in developing the minor including an analysis of competitive academic programs, evaluation of the current resources, qualifications and faculty considerations, the process of developing the program objectives and learning outcomes, and assessment strategies. The program will be run jointly by Criminal Justice and Computer Science departments, and faculty will communicate regularly to track the number of students in the minor and their progress through the curriculum. Both departments will ensure that the minor provides appropriate course content and learning experiences for graduates seeking employment. In our discussion, we will focus on challenges of designing the balanced curriculum to make it accessible for criminal justice and other non-computer science/computer information systems majors, the need of designing new courses and renovating existing courses to answer growing need to address this new emerging field. Lightning talk will also discuss the anticipated cost of the program, required resources, recruitment strategies, and the administrative approval mechanism.

Macrocyclic Compounds from Ansamycin Antibiotic Class as Inhibitors of PD1-PDL1 Protein-Protein Interaction.

The ability of tumors to escape from immune destruction is attributed to the protein-protein interaction between programmed cell death protein 1 (PD1) and programmed cell death ligand 1 (PDL1) proteins expressed by immune T cells and cancer cells, respectively. Therefore, pharmacological inhibition of the PD1-PDL1 interaction presents an important therapeutic target against a variety of tumors expressing PDL1 on their cell surface. Recently, five antibodies have been approved and several are in clinical trials against the PD1-PDL1 protein-protein interaction target. In contrast, there are very few reports of small-molecule inhibitors of PD1-PDL1 interaction, and most of them have relatively modest or weak inhibition activities, emphasizing the difficulty in designing small-molecule inhibitors against this challenging target. Therefore, we focused our attention on macrocycles that are known to exhibit target activity comparable to large macromolecules despite having molecular weights closer to small, drug-like molecules. In this context, our present study led to the identification of several macrocyclic compounds from the ansamycin antibiotics class to be inhibitors of PD1-PDL1 interaction. Importantly, one of these macrocyclic antibiotics, Rifabutin, showed an IC50 value of ca. 25 µM. This is remarkable considering it has a relatively low molecular weight and still is capable of inhibiting PD1-PDL1 protein-protein interaction whose binding interface spans over ca. 1970 Å2. Thus, these macrocycles may serve as guiding points for discovery and optimization of more potent, selective small-molecule inhibitors of PD1-PDL1 interaction, one of the most promising therapeutic targets against cancer.

A quadcopter controlled by brain concentration and eye blink

Brain computer interface (BCI) is a technology that enables a user to interact with the outside world by measuring and analysing signals associated with neural activity, and mapping an identified neural activity pattern to a behavior or action. In this work, an BCI system was developed where the operation of a quadcopter is controlled by identified brain concentration and eye blink patterns. A portable electroencephalography (EEG) headset is used to acquire neural signal around forehead and both eyes. Acquired EEG data are sent to a data processing computer wirelessly and processed in real-time. Identified brain concentration and eye blink patterns are associated with quadcopter operation commands and transmitted to the remote control that is modified to interface with the computer. The BCI system was evaluated by an experiment study and classification accuracy was calculated. Experimental results indicate that the system can achieve the expected performance without using EEG data from all channels and complicated data processing algorithms.