Suk Hoon Jung

Building a Practical Wi-Fi-Based Indoor Navigation System

This article presents the seven-step process involved in building a practical Wi-Fi-based indoor navigation system, which was implemented at the COEX complex in Seoul, Korea, in October 2010. The article describes the primary activities in each step using the COEX example. More than 200,000 users have downloaded the system since its first release. The successful launch of the COEX indoor navigation system suggests that indoor navigation is becoming a reality.

Protein complex prediction based on simultaneous protein interaction network

MOTIVATION The increase in the amount of available protein-protein interaction (PPI) data enables us to develop computational methods for protein complex predictions. A protein complex is a group of proteins that interact with each other at the same time and place. The protein complex generally corresponds to a cluster in PPI network (PPIN). However, clusters correspond not only to protein complexes but also to sets of proteins that interact dynamically with each other. As a result, conventional graph-theoretic clustering methods that disregard interaction dynamics show high false positive rates in protein complex predictions. RESULTS In this article, a method of refining PPIN is proposed that uses the structural interface data of protein pairs for protein complex predictions. A simultaneous protein interaction network (SPIN) is introduced to specify mutually exclusive interactions (MEIs) as indicated from the overlapping interfaces and to exclude competition from MEIs that arise during the detection of protein complexes. After constructing SPINs, naive clustering algorithms are applied to the SPINs for protein complex predictions. The evaluation results show that the proposed method outperforms the simple PPIN-based method in terms of removing false positive proteins in the formation of complexes. This shows that excluding competition between MEIs can be effective for improving prediction accuracy in general computational approaches involving protein interactions. AVAILABILITY http://code.google.com/p/simultaneous-pin/. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Unsupervised Learning for Crowdsourced Indoor Localization in Wireless Networks

Wireless Local Area Network (WLAN) location fingerprinting has become a prevalent approach to indoor localization. However, its widespread adoption has been hindered by the need for manual efforts to collect location-labeled fingerprints for the calibration of a localization model. Several semi-supervised learning methods have been applied to reduce such manual efforts by exploiting unlabeled fingerprints, but they still require some amount of labeled fingerprints for initializing the learning process. In this research, in order to obviate the need for location labels or references, we propose a novel unsupervised learning method that calibrates a localization model using unlabeled fingerprints based on a hybrid global-local optimization scheme. The method determines the optimal placement of fingerprint sequences on an indoor map, under the constraint imposed by the inner structure shown on the map such as walls and partitions. An efficient interaction between a global and a local optimization in the hybrid scheme drastically reduces the complexity of the learning task. Experiments carried out in a single- and a multi-story building revealed that the proposed method could successfully build a precise localization model without any location reference or explicit efforts to collect labeled samples.

A Computational Model for Predicting Protein Interactions Based on Multidomain Collaboration

Recently, several domain-based computational models for predicting protein-protein interactions (PPIs) have been proposed. The conventional methods usually infer domain or domain combination (DC) interactions from already known interacting sets of proteins, and then predict PPIs using the information. However, the majority of these models often have limitations in providing detailed information on which domain pair (single domain interaction) or DC pair (multidomain interaction) will actually interact for the predicted protein interaction. Therefore, a more comprehensive and concrete computational model for the prediction of PPIs is needed. We developed a computational model to predict PPIs using the information of intraprotein domain cohesion and interprotein DC coupling interaction. A method of identifying the primary interacting DC pair was also incorporated into the model in order to infer actual participants in a predicted interaction. Our method made an apparent improvement in the PPI prediction accuracy, and the primary interacting DC pair identification was valid specifically in predicting multidomain protein interactions. In this paper, we demonstrate that 1) the intraprotein domain cohesion is meaningful in improving the accuracy of domain-based PPI prediction, 2) a prediction model incorporating the intradomain cohesion enables us to identify the primary interacting DC pair, and 3) a hybrid approach using the intra/interdomain interaction information can lead to a more accurate prediction.

Elekspot: A Platform for Urban Place Recognition via Crowdsourcing

The proliferation of Wi-Fi infrastructures has facilitated numerous indoor localization techniques using Wi-Fi location fingerprints. They make it possible to identify a room or a place in urban environment, which is especially important in enabling many interesting location-based services. As there are too many rooms and places such as cafes and restaurants to be recognized in urban environment, the crowdsourcing approach has been proposed to collect Wi-Fi location fingerprints based on user participation. However, its actual deployment in a large-scale urban environment presents numerous design and implementation challenges due to urban characteristics such as a large crowd, dense region, and device diversity. This paper presents Elekspot system, whose design goal is to support system scalability, device heterogeneity, robustness against lack of contributions, and localization accuracy. Through several experiments and implementation of actual applications targeting urban places we confirmed that the architecture and methods of Elekspot can effectively meet the design goals.

Performance Evaluation of Radio Map Construction Methods for Wi-Fi Positioning Systems

A radio map is a collection of signal fingerprints labeled with their collected locations. It is known that the performance of a fingerprint-based positioning systems is closely related to the precision and accuracy of the underlying radio maps. However, little has been studied on the performance of radio maps in relation to the fingerprint collection methods and the radio map models, which determine the accuracy and precision of radio maps, respectively. This paper evaluates the performance of various radio map construction methods in both indoor and outdoor environments. Four radio map construction methods, i.e., a point-by-point manual calibration, a walking survey, a semisupervised learning-based method, and an unsupervised learning-based method, have been compared. We also evaluate the performance of various types of radio map models that represent the characteristics of collected fingerprints. To demonstrate the importance of the radio map model, a new model named signal fluctuation matrix (SFM) was developed, and its performance was compared with that of the three conventional radio map models, respectively. The evaluation revealed that the performance of the radio maps was very sensitive to the design of radio map models and the number of fingerprints collected at each location. The performance achieved by SFM-based positioning was comparable with that of the other models despite using a small number of fingerprints.

A Multi-Classifier Approach for WiFi-Based Positioning System

WLAN fingerprint-based positioning systems are a viable solution for estimating the location of mobile stations. Recently, various machine learning techniques have been applied to the WLAN fingerprint-based positioning systems to further enhance their accuracy. Due to the noisy characteristics of RF signals as well as the lack of the study on environmental factors affecting the signal propagation, however, the accuracy of the previously suggested systems seems to have a strong dependence on numerous environmental conditions. In this work, we have developed a multi-classifier for the WLAN fingerprint-based positioning systems employing a combining rule. According to the experiments of the multi-classifier performed in various environments, the combination of the multiple numbers of classifiers could significantly mitigate the environment-dependent characteristics of the classifiers. The performance of the multi-classifier was found to be superior to that of the other single classifiers in all test environments; the average error distances and their standard deviations were much more improved by the multi-classifier in all test environments.

A crowdsourcing-based global indoor positioning and navigation system

This paper presents three key techniques to realize a global indoor positioning system (GIPS), and a global in- and-outdoor integrated navigation system (GINS). A crowdsourcing radio map construction method, a positioning algorithm for crowdsourced radio maps, and an indoor and outdoor environment detection method are developed as the three key techniques. The developed techniques have been integrated into a crowdsourcing-based indoor positioning system, named KAILOS, aiming to realize the GIPS and GINS. The system was deployed at KAIST, Daejeon campus, and now an in- and-outdoor integrated navigation service is available at KAIST campus area. The successful launch of KAILOS foretells that the GIPS and GINS are becoming a reality.

Identification of Conserved Domain Combinations in S.cerevisiae Proteins

In this paper, we propose a formulated method for the analysis of conserved domain combinations and report an overview of domain combinations by identifying domain patterns and analyzing their functional annotations. The proposed method measures co-occurrence frequency and mutual dependency of domains in a domain combination using association rules. The method is useful to estimate the meaningfulness of a given domain combination in terms of conservation. Using the method, we extracted domain patterns in S.cerevisiae proteins and investigated GO term annotations of the domains. According to the investigation, domains in S.cerevisiae proteins are turned out to form patterns in which the members of the patterns are highly affiliated to one another. Also, extracted patterns are revealed to have a tendency of being associated with molecular functions.

Uncaught signal imputation for accuracy enhancement of WLAN-based positioning systems

In this paper we propose a technique to enhance the accuracy of WiFi fingerprint-based localization by imputing uncaught access point (AP) signals of WiFi fingerprints. Two techniques were developed for this; one is to impute uncaught AP signals by referring to WiFi radio map (WRM) fingerprints at the very previous location, another is referring to WRM fingerprints obtained by predicting the next location. When we measured the accuracy of localization at an E-Mart, Seoul, Korea and a KAIST Library, Daejeon, Korea with and without uncaught signal imputation, the imputed signal resulted in around 30% better accuracy improvement than the signals without imputation. In addition, the imputation methods using WRM information showed significantly better accuracy than using a fixed value for uncaught AP signals. This indicates that the uncaught signal imputation, which was overlooked in the WLAN-based localization, should be incorporated with other filtering techniques for a more reliable and accurate localization.