Young-Tack Park

A novel feature selection method based on normalized mutual information

In this paper, a novel feature selection method based on the normalization of the well-known mutual information measurement is presented. Our method is derived from an existing approach, the max-relevance and min-redundancy (mRMR) approach. We, however, propose to normalize the mutual information used in the method so that the domination of the relevance or of the redundancy can be eliminated. We borrow some commonly used recognition models including Support Vector Machine (SVM), k-Nearest-Neighbor (kNN), and Linear Discriminant Analysis (LDA) to compare our algorithm with the original (mRMR) and a recently improved version of the mRMR, the Normalized Mutual Information Feature Selection (NMIFS) algorithm. To avoid data-specific statements, we conduct our classification experiments using various datasets from the UCI machine learning repository. The results confirm that our feature selection method is more robust than the others with regard to classification accuracy.

Human Facial Expression Recognition Using Stepwise Linear Discriminant Analysis and Hidden Conditional Random Fields

This paper introduces an accurate and robust facial expression recognition (FER) system. For feature extraction, the proposed FER system employs stepwise linear discriminant analysis (SWLDA). SWLDA focuses on selecting the localized features from the expression frames using the partial F-test values, thereby reducing the within class variance and increasing the low between variance among different expression classes. For recognition, the hidden conditional random fields (HCRFs) model is utilized. HCRF is capable of approximating a complex distribution using a mixture of Gaussian density functions. To achieve optimum results, the system employs a hierarchical recognition strategy. Under these settings, expressions are divided into three categories based on parts of the face that contribute most toward an expression. During recognition, at the first level, SWLDA and HCRF are employed to recognize the expression category; whereas, at the second level, the label for the expression within the recognized category is determined using a separate set of SWLDA and HCRF, trained just for that category. In order to validate the system, four publicly available data sets were used, and a total of four experiments were performed. The weighted average recognition rate for the proposed FER approach was 96.37% across the four different data sets, which is a significant improvement in contrast to the existing FER methods.

Ontology-based multi-layered robot knowledge framework (OMRKF) for robot intelligence

An ontology-based multi-layered robot knowledge framework (OMRKF) is proposed to implement robot intelligence to be useful in a robot environment. OMRKF consists of four classes of knowledge (KClass), axioms and two types of rules. Four KClasses including perception, model, activity and context class are organized in a hierarchy of three knowledge levels (KLevel) and three ontology layers (OLayer). The axioms specify the semantics of concepts and relational constraints between ontological elements in each OLayer. One type of rule is designed for relationships between concepts in the same KClasses but in different KLevels. These rules will be used in a way of unidirectional reasoning. And, the other types of rules are also designed for association between concepts in different KLevels and different KClasses to be used in a way of bi-directional reasoning. These features will let OMRKF enable a robot to integrate robot knowledge from levels of sensor data and primitive behaviors to levels of symbolic data and contextual information regardless of class of knowledge. To show the validities of our proposed OMRKF, several experimental results will be illustrated, where some queries can be possibly answered by using uni-directional rules as well as bi-directional rules even with partial and uncertain information.

Scalable OWL-Horst ontology reasoning using SPARK

In this paper, we present an approach to perform reasoning for scalable OWL ontologies in a Hadoop-based distributed computing cluster. Rule-based reasoning is typically used for a scalable OWL-Horst reasoning; typically, the system repeatedly performs many operations involving semantic axioms for big ontology triples until no further inferred data exists. Thus, the reasoning systems suffer from performance limitations when ontology reasoning is performed via disk-based MapReduce approaches. To overcome this drawback, we propose an approach that loads triples to memory in computer nodes that are connected by SPARK - a memory-based cluster computing platform - and executes ontology reasoning. To implement an OWL Horst ontology reasoning system, we first define a set of algorithms such that they divide large triples into Resilient Distributed Datasets (RDDs), taking into account the patterns and interdependencies of the reasoning rules. We then load each RDD into the memory of computers composing a distributed computing cluster and subsequently perform distributed reasoning by rule execution orders. To evaluate the proposed methods, we compare it to WebPIE using the LUBM set, which is formal dataset for evaluating ontology inferences and search speeds. The proposed approach shows throughput is improved by 200% (98k/sec) as compared to WebPIE (33k/sec) using the LUBM6000 (860 million triples, 109 gigabyte).

Real-time and Parallel Semantic Translation Technique for Large-Scale Streaming Sensor Data in an IoT Environment

Nowadays, studies on the fusion of Semantic Web technologies are being carried out to promote the interoperability and value of sensor data in an IoT environment. To accomplish this, the semantic translation of sensor data is essential for convergence with service domain knowledge. The existing semantic translation technique, however, involves translating from static metadata into semantic data(RDF), and cannot properly process real-time and large-scale features in an IoT environment. Therefore, in this paper, we propose a technique for translating large-scale streaming sensor data generated in an IoT environment into semantic data, using real-time and parallel processing. In this technique, we define rules for semantic translation and store them in the semantic repository. The sensor data is translated in real-time with parallel processing using these pre-defined rules and an ontology-based semantic model. To improve the performance, we use the Apache Storm, a real-time big data analysis framework for parallel processing. The proposed technique was subjected to performance testing with the AWS observation data of the Meteorological Administration, which are large-scale streaming sensor data for demonstration purposes.

Semantic sleep management service in healthcare sensor networks

In this paper, we propose a semantic sleep management service using healthcare sensors(blood pressure, blood sugar, body temperature, snoring, sleep apnea) and private health information(age, gender, weight, smoking amount and drinking quantity). The proposed service finds the best private sleep pattern by acquiring sensor observations, providing analysis results of private sleep trend by analyzing data gathered from a heterogeneous individual healthcare sensors and private health information in order to enhance sleep quality of each individual. To this end, we use newly-made sleep management sensor to detect the snoring time and the number of sleep apnea and use semantic web technologies to represent standard specification and processing of sensor networks using ontologies that allow representation of structural properties of event types and constraints between them.

OnCU system: ontology-based category utility approach for author name disambiguation

Author name disambiguation is essential for improving performance of document indexing, retrieval, and web search. Author name disambiguation resolves the conflict when multiple authors share the same name label. This paper introduces a novel approach which exploits ontologies and ontology-based category utility for author name disambiguation. Author name disambiguation determines the correct author from various candidate authors in the populated author ontology. Candidate authors are evaluated using proposed ontology-based category utility to resolve disambiguation. Ontology-based category utility has been proposed to exploit semantic information in ontology for semantic analysis for disambiguation. The ontology-based category utility increases the number of disambiguation by about 10% compared with that of category utility, and increases the overall amount of accuracy by around 98%.

Enhanced Search Method for Ontology Classification

The Web ontology language (OWL) has become a W3C recommendation to publish and share ontologies on the semantic web. In order to derive hidden information (classification, satisfiability and realization) of OWL ontology, a number of OWL reasoners have been introduced. Most of reasoners use both top-down and bottom-up search for ontology classification. In this paper, we propose an enhanced method of optimizing the ontology classification process of ontology reasoning. One goal of this paper is to provide such a available algorithm for future implementers of ontology reasoning system. Building the optimization method that came off best into ontology reasoning system greatly enhanced its efficiency. Our work focuses on two key aspects: The first and foremost, we describe classical methods for ontology classification. As subsumption testing to classify ontology is costly, it is important to ensure that the classification process uses the smallest number of tests. Therefore, we consider enhanced method and evaluate their effect on four different types of test ontology. The result of the experiment was that the enhanced search method increases performance improvement 30% something like that compare with the classical method.

Large-scale incremental OWL/RDFS reasoning over fuzzy RDF data

Ontological RDF data are extracted from multiple sources on the web through mapping and alignment for various purposes, but extracting and reasoning about ontologies from different sources causes information ambiguity and uncertainty. A reasonable solution to this problem is to annotate extracted ontology data with truth values to determine the reliability of information. However, the recent growth in data has brought forth difficulties in ascertaining the credibility of numerous ontologies during OWL/RDFS reasoning. In this paper, we present a distributed and incremental reasoning approach for RDF data with uncertainty. We focused on RDFS and OWL pD* semantics and developed methods for incremental OWL reasoning with uncertainty. We also introduced parallel algorithms that resolve the scalable reasoning problem. To evaluate the efficiency of the proposed system, we conducted OWL/RDFS reasoning over fuzzy LUBM3000 and achieved a performance three times higher than that achieved with the fastest reasoning system.

Scalable Ontology Reasoning Using GPU Cluster Approach

In recent years, there has been a need for techniques for large-scale ontology inference in order to infer new knowledge from existing knowledge at a high speed, and for a diversity of semantic services. With the recent advances in distributed computing, developments of ontology inference engines have mostly been studied based on Hadoop or Spark frameworks on large clusters. Parallel programming techniques using GPGPU, which utilizes many cores when compared with CPU, is also used for ontology inference. In this paper, by combining the advantages of both techniques, we propose a new method for reasoning large RDFS ontology data using a Spark in-memory framework and inferencing distributed data at a high speed using GPGPU. Using GPGPU, ontology reasoning over high-capacity data can be performed as a low cost with higher efficiency over conventional inference methods. In addition, we show that GPGPU can reduce the data workload on each node through the Spark cluster. In order to evaluate our approach, we used LUBM ranging from 10 to 120. Our experimental results showed that our proposed reasoning engine performs 7 times faster than a conventional approach which uses a Spark in-memory inference engine.