Querying Patterns in High-Dimensional Heterogenous Datasets

Singh, Vishwakarma University of California, Santa Barbara 2012 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The recent technological advancements have led to the availability of a plethora of heterogenous datasets, e.g., images tagged with geo-location and descriptive keywords. An object in these datasets is described by a set of high-dimensional feature vectors. For example, a keyword-tagged image is represented by a color-histogram and a word-histogram. Analyzing these datasets gives better insights into the processes generating the datasets, opens new frontiers of scientific research, and fuels development of life-changing products. An effective mechanism for exploring these heterogenous datasets is querying. One such kind of query is a pattern query. Given a heterogenous dataset and a query, the task here is to find a set of objects which are constrained by a relationship and satisfy the query. For example, given a dataset of keyword-tagged objects, a useful pattern query is to find a set of similar objects that contains a given set of keywords. Querying patterns in high-dimensional heterogenous datasets brings about a new set of computational challenges. High performance algorithms to efficiently and accurately query patterns are presented in this thesis. First, a scalable algorithm, SIMP, is described for accurately querying near neighbors in a high-dimensional dataset. SIMP significantly outperforms the state-of-the-art techniques. Next, a novel algorithm, ProMiSH, is proposed for efficiently querying patterns by keywords. ProMiSH has a speed-up of more than four orders over the state-of-the-art techniques. Then, an algorithm, QUIP, is described for querying patterns by example in a spatial dataset, e.g., geographical maps. QUIP offers an improvement of 87% in running time over the baseline approach. Next, an algorithm for querying patterns by example in a temporal dataset is described. It specifically solves the problem of finding duplicate videos. The proposed algorithm yields a practical query time for video duplicate detection. Finally, a scalable method to compute statistical significance of results of a multi-object query is discussed. Statistical significance or p-value provides a more useful criterion for ranking the results of a query.

Integration of the visual features for content-based image retrieval : Kulwinder Singh

Singh, Kulwinder Graduate School of PaiChai University 2003 국내박사

Expression des neurotrophen Faktors CNTF (Ciliary Neurotrophic Factor) unter der Kontrolle eines neuronenspezifischen Promotors

Singh, Guido 1993 국내박사

Successive stiffness increment and successive force augment approach for enhancing haptic and teleoperation systems

Singh, Harsimran Korea University of Technology and Education 2017 국내박사

There have been numerous approaches proposed to enlarge the impedance range of haptic interaction, however most of these approaches sacrifice the actual displayed stiffness at the cost of stability. It is also a challenging issue to increase the rate-hardness of haptic interaction while maintaining stability. The actual perceived rate-hardness has been much lower than what physical systems have, and it hasn’t been enough to satisfy what users expect to feel. A novel approach, Successive Stiffness Increment (SSI), which successively increases the stiffness as the number of interaction cycle increase, is proposed. The stiffness is sequentially modulated from a low value to a high value, close to the desired stiffness while maintaining stability. This sequential stiffness increment was possible because the proposed approach guarantees the convergence of the penetration distance and increases the feedback force with every successive interaction cycle. Another similar approach, Successive Force Augment (SFA), is proposed which not only enlarges the impedance range but also enhances the rate-hardness of haptic interaction. Both the approaches were evaluated using commercially available Phantom Premium 1.5 and Phantom Omni. The experimental results and comparison with other conventional approaches showed that the proposed approach delivered a much higher displayed stiffness and achieved increased rate-hardness. Enhancing transparency of a teleoperation system by increasing the command following bandwidth has not received lots of attention so far. This is considered a challenging task since in a teleoperation system the commandfollowing bandwidth of the slave robot motion controller cannot be increased with a conventional motion controller as the desired trajectory is instantaneously commanded by the human operator and thus, cannot be considered to be given in a pre-computed, smooth second order derivative form. Therefore, I propose a method to increase the command-following bandwidth by extending the previously introduced Successive Stiffness Increment approach to bilateral teleoperation. The approach allows realizing a very high motion controller gain, which cannot be realized with a conventional bilateral teleoperation controller as confirmed by experimental results. The Successive Force Augment approach was also extended to time-delayed teleoperation systems to ensure robust stability and to increase its transparency. Experimental evaluation and comparison with Time Domain Passivity Approach (TDPA) shows that the proposed framework provides a more transparent system.

Machine Learning Based Indoor Localization Using Wi-Fi RSSI Fingerprints

Singh, Navneet 가톨릭대학교 대학원 2021 국내석사

Localization technologies have been developed extensively over the last decade by using technological advancement in wireless and computation devices to provide location and navigation services to its users. Humans spent most of their time indoors than outdoor, therefore localization systems specifically developed for indoor environments. Extensive research has been done to develop indoor localization systems using several signal technologies such as Bluetooth, ZigBee, UWB, Wi-Fi, etc., depending upon context and application requirements. From the last decade, the reach of Wi-Fi enabled devices to increase exponentially, from our homes to large areas like airports and stadiums. Thus Wi-Fi-based indoor localization system could become an integrated part of the internet of things in the indoor environment. Moreover, Wi-Fi is compatible with various parameters such as RSSI, AoA, ToF, and CSI. Moreover, RSSI is the most simple parameter used as the basis for the machine learning (ML)-based fingerprinting localization approach. This master thesis firstly explores the various aspects (RSSI, Wi-Fi fingerprinting, database construction, enrichment techniques and publicly available datasets) of ML-based fingerprinting localization using Wi-Fi RSSI, which act as a literature review for our proposed scheme. Secondly, XBGLoc: PCA-XGBoost based ML model is proposed to localize the target device with high accuracy in multi-building multi-floor indoor settings. Finally, we evaluate the proposed model on various datasets such as the UJIIndoor dataset, Tampere dataset, and Alcala dataset and compare the performance with various schemes. These evaluations show that XGBLoc is scalable, robust and can be applied to real-world applications. Also, discuss various application scenarios and open challenges of ML-based fingerprinting localization. 위치탐지 방안은 사용자에게 위치 및 내비게이션 서비스 제공을 위한 근간기술로 4차 산업을 타깃으로 한 무선 및 컴퓨팅 장치의 기술 (인공지능, 5G/6G 등)의 발전으로 지난 10 년 동안 여러 진전이 있어 왔다. 사람은 대부분의 시간을 실외보다는 실내에서 보낸다. 본 논문에서는 이러한 옥내 환경에서 요구되는 새로운 위치탐지 시스템 설계방안을 강구한다. 현재까지 대부분의 무선통신 시스템은Wi-Fi, Bluetooth, ZigBee, UWB 등 다양한 표준 무선 방식을 사용해왔으며 옥내 위치 탐지 기술 또한 이러한 표준 방식을 이용한 여러 연구가 행해지고 있다. 특별히 여타 표준 무선 방식에 비해Wi-Fi의 경우, 서비스 커버리지 영역이 점점 커지고 있으며 이에 비례하여 이용 서비스 모바일 기기 수 또한 급속도로 증가하고 있다. Wi-Fi 기반의 정밀 실내 위치 탐지 방안은 사물 인터넷 기반의 미래형 응용 서비스를 고려 점점 시스템에 통합 설계되는 추세이다. Wi-Fi 기반 위치탐지를 위해 RSSI, AoA, ToF 및 CSI와 같은 다양한 매개 변수를 이용한 방식이 설계되어 왔으며 그 중에서도 RSSI 방식은 상대적으로 구현이 간단한 반면에서 탐지성능은 떨어진다. 최근 들어 ML 기반의 RSSI 지문 프린팅 위치탐지 기술 연구가 RSSI 방식의 이러한 단점을 보완하는 새로운 대안으로 강구되고 있다. 본 석사 논문에서는 첫째, ML 기반 RSSI 지문 프린팅 위치 탐지 기술의 다양한 측면 (RSSI, Wi-Fi 지문, 데이터베이스 구성, 강화 기술 및 공개적으로 사용 가능한 데이터 세트 등)에 대한 조사를 포함한다. 여기에는 최근 문헌에 나와 있는 Wi-Fi RSSI를 이용한 인공지능 기반의 다양한 위치탐지 방안을 우선 살펴본다. 둘째, 본 논문에서 제안한 새로운 Wi-Fi RSSI를 이용한 인공지능 위치탐지 방안인 XBGLoc : PCA-XGBoost 기반 ML 모델을 설계하고 다중 빌딩 다층 실내 환경에서 높은 정확도로 타깃 장치에 대한 위치탐지가 가능함을 보여준다. 마지막으로, 제안한 위치 탐지 모델을 여러 공개RSSI DB(데이터 베이스), 즉, UJIIndoor 데이터 셋, Tampere 데이터 셋, Alcala 데이터 셋에 각각 적용하여 시뮬레이션하고 기존문헌에서 제안한 여타 다양한 위치탐지 방식과도 성능을 비교한다. 이러한 비교평가를 통해 제안방식인 XGBLoc이 상대적으로 확장성이 높으며, 무선채널에 견고하고, 실용성 면에서도 우수함을 보여준다. 아울러, ML 기반 지문 프린팅 위치탐지의 다양한 애플리케이션 시나리오를 알아보고 향후 해결해야 할 숙제에 대해서도 간략히 논한다.

A virtual user agent for location-based global access to directory services using atom for automatic update

Singh, Kulwinder Paichai University 2008 국내박사

Synthesis of azasugars and ceramide analogs from Chiral Aziridine-2-carboxylate

Singh, Alok 韓國外國語大學校 大學院 2011 국내박사

The importance of azasugars has grown over the last 50 years, particularly in the pharmaceutical industries. Therefore the developments of new methods for synthesis of azasugars are a very attractive research area. In Chapter 1, Section A, the properties and impact of the azasugars in organic synthesis are overviewed. Existing stereoselective dihydroxylation methods are reviewed and new developments in asymmetric synthesis of azasugars are discussed. As asymmetric synthesis of 1-deoxyazasugars from chiral aziridines, Piperidine azasugars natural products including deoxynojirimycins were isolated from plants and microorganism with many valuable biological activities including potent glycosidase inhibition. Those were synthesized starting from chiral aziridine-2-carboxylate on the basis of the common synthetic intermediate bicyclic oxazolidinyl-1, 2, 3, 6-tetrahydropyridine1,2 originated from the RCM. In Section B, described the synthesis of OGT 2378. That is a novel and potent anticancer agent that inhibits the synthesis of ganglliosides in cancer cells with no cytotoxic or antiproliferative effects. A practical and highly enantioselective synthesis of OGT2378 has been achieved for the first time using the aziridine-2-carboxylate as starting material and as the source of chirality. The key step in synthesis is reductive alkylation of deoxyidonojirimycin derivative. With the advancement in synthetic strategies proper tailoring of stereo, regioselectivity of biologically active compounds are possible. In Section C, described the formal synthesis of fagomine. Fagomines are a family of naturally occurring 1, 2-dideoxyazasugars which have varying levels of oxygenation and a stereochemical array on the piperidine ring. We envisioned dioxanylpiperidene as a new chiral building block, which represents an ideal precursor for synthesis of fagomine and its epimer. In Section C, described the stereoslective synthesis of (2R, 3S)-2-hydroxymethylpiperdin-3-ol via the region-selective and diastereoselective introduction of ring opening of aziridine ring and ring-closing metathesis. 3-hydroxy-2-hydroxymethyl piperidine has a right stereochemistry for synthesis of Febrifugine which is hydrangea alkaloid that possesses antimalerial and anticoccidial properties. The synthetic strategy described has significant potential for further extension to other isomers and related analogues. In Chapter 2, Ceramide is an apoptosis-inducing signaling molecule in the sphingolipid metabolism which emerges as a potential target for anticancer therapy. I have prepared a series of ceramide analogs starting from aziridine-2-carboxylate to make oxazole moiety as common synthetic intermediate. Constrained ceramide analogs were designed and synthesized by binding terminal alcohol and amine of ceramide with additional functional group in alkene chain via cross-metathesis in high yield.

William Carey's missionary methods and their relevance to the Indian church

Singh, Wijai Kumar Presbyterian Theological Seminary 1980 국내박사

Colonial state, women victims and criminals : the north-western provinces and oudh, (india), 1870-1910

Singh, Priyam Bada Temple University 1993 국내박사

LIFE STYLE AND PSYCHOGRAPHICS

Singh, Sushil Kumar 漢陽大學校 大學院 1984 국내석사