A Method of Reducing Dimensionality of Dissimilarity for Pattern Recognition

Gao Jian 명지대학교 일반대학원 2008 국내석사

요약 논문제목: 패턴인식을 위한 비유사도표현의 차원축소방법 본 논문에서는 비유사도기반 식별기(DBC)를 효율적으로 구현할 수 있는 새로운 방법을 연구하였다. DBC는 Duin교수팀에 의해 제안된 식별 방법으로, 패턴을 구성하는 특징으로 식별하는 대신에 패턴들 사이의 비유사도를 측정하여 클래스를 분류하는 방법이다. DBC 식별기를 설계하기 위해서는 전체 학습 데이터를 대표할 수 있는 적은 규모의 프로토타입을 추출하여야 한다. 그런데 최적의 프로토타입을 선정하기가 어렵고, 또한 선정과정에서 식별에 유용한 정보를 상실할 수도 있다. 따라서 본 연구에서는 전체 학습 데이터를 프로토타입으로 선정한 다음, 차원축소 방법(DRS)을 적용하는 새로운 방법을 제안하였다. 즉, 학습 데이터에서 소수의 프로토타입을 추출하여 이용하는 대신에 전체 데이터로 비유사도 행렬을 작성한 다음 기존의 DRS를 적용하여 차원을 축소하는 방법이다. 두 종류의 실생활 벤치마크 데이터베이스로 실험한 결과, 제안방법은 기존의 방법에 비하여 식별율을 개선하였다. Abstract Title: A Method of Reducing Dimensionality of Dissimilarity for Pattern Recognition Author: Jian Gao Supervisor: Professor Sang-Woon Kim, PhD The aim of this thesis is to present a strategy by which a new philosophy for pattern classification pertaining to dissimilarity-based classifiers (DBCs) can be efficiently implemented. Proposed by Duin and his co-authors, DBCs are a way of defining classifiers among classes; they are not based on the feature measurements of individual patterns, but rather on a suitable dissimilarity measure among the patterns. The problem with this strategy is that we need to select a representative set of data that is both compact and capable of representing the entire data set. However, it is difficult to find the optimal number of prototypes and, furthermore, selecting prototype stage may potentially lose some useful information for discrimination. To avoid these problems, in this thesis, we propose an alternative approach where we use all available samples from the training set as prototypes and subsequently apply dimensionality reduction schemes. That is, we prefer not to directly select the representative prototypes from the training samples; rather, we use a dimensionality reduction scheme after computing the dissimilarity matrix with the entire training samples. Our experimental results demonstrate that the proposed mechanism can improve the classification accuracy of conventional approaches for two real-life benchmark databases.

Sustainable charged materials for stable triboelectric nanogenerator performance

Gao, Jian Sungkyunkwan University 2024 국내석사

Applying sustainable charged species to triboelectric nanogenerators (TENGs) is important to improve performance and sustainability. Triboelectric nanogenerators are devices that use tiny mechanical movements generated by friction to collect energy, with charged bodies playing a central role. First, the use of sustainably charged materials can help improve the efficiency of power generation. Traditional charged materials such as polytetrafluoroethylene (PTFE) can produce static electricity to a certain extent but may contain environmentally unfriendly chemicals during manufacturing and handling. In contrast, sustainable charged materials, such as biodegradable polymers or new materials based on renewable resources, are more in line with environmental protection concepts and help reduce the environmental burden during the preparation process. Second, using sustainably charged materials helps extend the life of the triboelectric nanogenerator. Due to the good durability and stability of sustainable materials, they can withstand prolonged frictional movements better than conventional materials without losing the charge effect. This is especially important for some application scenarios that require continuous practice and use in various environments. Here, we show that poly(2-vinylpyridine) (P2VP) (5-20 wt%) and titanium-phase barium oxide (BaTiO3) can be chemically mixed to form hyperdispersed core-shell impurity materials (P2VP@BaTiO3), forming high uniform film of quality. Triboelectric nanogenerators (TENG) based on P2VP@BaTiO3, especially those based on 10wt% P2VP (P2VP10@BaTiO3), have greatly improved output performance compared with similar products of physically mixed P2VP/BaTiO3. The double-layer TENG based on P2VP10@BaTiO3 not only has a good transport charge density of 281.7 μC×m^(-2) and a power density of 27.2 〖W×m〗^(-2), but also has excellent component stability (the maximum output voltage after 54,000 cycles can last for approximately 100%), the voltage retention rate is approximately 68.7% at 99% humidity. In particular, if a MoS2/SiO2/Ni Mesh layer is introduced into a double-layer TENG, the charge density can be further improved to the highest 1228 μC×m^(-2) reported for a TENG so far. Additionally, we demonstrate a near-field communication-based sensing system for monitoring carbon dioxide gas using a self-powered generator with improved output performance and robustness.

Synthesis and characterization of two dimensional transition metal dichalcogenides

Gao, Jian Rensselaer Polytechnic Institute 2016 해외박사(DDOD)

Two-dimensional transition metal dichalcogenides (TMDs) are an emerging class of atomically thin semiconductors that show potential in next-generation electronics, optoelectronics, and energy storage batteries. The successful synthesis and doping of TMDs is the key to their applications. I have synthesized monolayer MoS2, WS2, and multilayer ReS2 flakes by CVD, and studied an unprecedented one-pot synthesis for transition-metal substitution doping in large-area, synthetic monolayer TMDs. Electron microscopy, optical and electronic transport characterization and ab initio calculations indicate that our doping strategy preserves the attractive qualities of TMD monolayers, including semiconducting transport and strong direct-gap luminescence. The Re doping of MoS2 greatly improve the contact quality (one of the biggest issue in TMDs) and the FET shows Ohmic contact even at low temperature (4K). These results potentially enables next-generation optoelectronic technology in the atomically-thin regime. Besides, TMDs are generally considered to be 'air-stable', however, we have found that they exhibit poor long-term stability in air in morphology, chemical states, photo-emission, and demonstrated a potential solution to this problem by encapsulation of the monolayer sheet with transparent parylene C. Synthetic TMDs tend to grow parallel to the growth substrate, however, high performance energy conversion and storage devices prefer flakes with high exposed surface area. Therefore by choosing the right precursors and appropriate tuning of the CVD growth conditions, we have grown vertical ReS2 nanosheets on various growth substrates. We show that these structural features of the vertically grown ReS2 sheets can be exploited to significantly improve their performance as electrochemical catalysts in Lithium-Sulfur (Li-S) batteries and in hydrogen evolution reactions (HER). After 300 cycles, the specific capacity of the Li-S battery with vertical-ReS2 catalyst is retained above 750 mA h g-1 with only ∼0.063% capacity decay per cycle, much better than the baseline battery. As a HER catalyst, the vertical-ReS 2 provides very small onset over-potential (< 100 mV) and an exceptional exchange current density (∼67.6 microA/cm2), which is superior to most previous work with TMD catalysts.

Multiplicative multifractional modeling of long-range-dependent (LRD) traffic in computer communications networks

Gao, Jian-Bo University of California, Los Angeles 2000 해외박사(DDOD)

Source traffic streams as well as aggregated traffic flows often exhibit long-range-dependent (LRD) properties. In this work. we model traffic streams using multiplicative multifractal processes. We develop two type of models, the multifractal point processes and multifractal counting processes. We prove a number of properties of multiplicative multifractal processes that are most relevant to their use as network traffic models. In particular, we prove that these processes also possess LRD property. We then choose the models to contain only one or two parameters. We show that the model has well defined burstiness descriptors, and is easy to construct. We consider a single server queueing system which is loaded, on one hand, by the measured processes, and, on the other hand, by properly parameterized multifractal processes. In comparing the system-size tail distributions, we demonstrate our model to effectively track the behavior exhibited by the system driven by the actual traffic processes. We also study the superposition of LRD flows. We prove a number of useful mathematical relations for the superimposed process. In particular, we prove that the superposition of a finite number of multiplicative multifractal traffic streams results asymptotically in another multifractal stream. Furthermore we demonstrate numerically that the superimposed process can be effectively modeled by an ideal multiplicative process. The burstiness behavior of the superimposed process is also studied, and is shown to be always less bursty than the most bursty source component.

(A) Comparative study of digital quality management level of manufacturing : Korea and China

Gao, Jianxi Sungkyunkwan University 2023 국내석사

In the era of digital economy, building an internationally competitive manufacturing industry with intelligent manufacturing as the main direction is the only way to transform the country into a manufacturing power and achieve high-quality economic development. The deep integration of digital technology and manufacturing has readjusted the international economic structure and triggered a new round of industrial transformation. The digital quality management of manufacturing is in the critical period of transformation and development. Understanding the self-positioning of enterprises and continuous improvement are conducive to the effective and efficient work of enterprises. This study compares existing quality management models through previous studies and national standards. Based on the maturity level of the DQM model, 5 dimensions, 15 quality criteria and 36 criterion parts, the digital capabilities of quality management in Korean and Chinese manufacturing enterprises were investigated, discussed the differences and characteristics of the digital quality management level of the two countries. Help enterprises of both countries to recognize the advantages and improvement opportunities of digital quality management, and further enhance quality competitiveness. 디지털 경제 시대에 지능형 제조업을 주요 방향으로 하여 국제적으로 경쟁력 있는 제조업을 구축하는 것만이 국가를 제조업 강국으로 전환시키고 고(高) 품질 경제 발전을 이룰 수 있는 유일한 방법이다. 디지털 기술과 제조업의 깊은 통합은 국제 경제 구조를 재조정하고 새로운 산업 변혁을 촉발했다. 제조업의 디지털 품질관리는 변화와 발전의 중요한 시기에 있다. 기업의 자기 위치 파악과 지속적인 개선은 기업의 효과적이고 효율적인 업무에 도움이 된다. 본 연구는 선행연구와 국가표준을 통해 기존의 품질관리 모델을 비교한다. DQM 모델의 성숙도를 바탕으로 한국과 중국 제조기업의 품질관리 디지털 역량을 5 가지 품질 범주, 15 개 품질기준, 36 개 품질 진단 소항목을 조사하고 양국의 디지털 품질관리 수준의 차이와 특성에 대해 논의하였다. 양국 기업이 디지털 품질관리의 특징과 개선 기회를 인식하고 품질 경쟁력을 더욱 강화할 수 있도록 지원한다.