http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : Gyuseong Lee (검색결과 2 건)
-
Lee, Gyuseong Sungkyunkwan university 2019 국내석사
As technology has evolved, demand of energy per person has continued to increase. Accordingly, a number of research teams have been investigating energy storage mechanisms. There has been intensive attention in the field of lithium ion batteries in decades because of its high potential performance on energy storage for electric vehicles and portable devices. Recently, most of these portable devices and EVs are using LIBs due to its high energy density and cost efficiency. However, present commercial lithium ion batteries which using graphite as anode material could not meet energy demand of electronic devices anymore. To satisfy the high demand of energy, development of new electrode material is indispensable. In this point of view, conversion reaction metal oxides are good candidates of graphite. Generally, conversion reaction metal oxides have high capacity and high energy density which proper for anode material of lithium ion battery. Nevertheless, conversion reaction metal oxides suffer from some disadvantages like low electronic conductivity, low retention, etc. Therefore, some special strategy is needed. Introduction of mesoporous structure could be one of strategies. Its high pore volume can buffer volume expansion of conversion reaction metal oxide. Also, shorten lithium ion diffusion length could increase electronic conductivity. Regarding the templates used for introducing mesoporous structure, we suggested that using colloidal silica as template of mesoporous material. This method is cost efficient and has straightforward way to synthesize mesoporous material. Thus, it is suitable for mass production in industry field. Moreover, hybridized metal oxide can show special features. Excess lithium storage compares with theoretical capacity and enhanced electrochemical performance would be implemented. In this work, we synthesized colloidal silica templated metal oxides and tested as anode material in lithium ion batteries with high capacity and good rate capability.
-
Exploring Guidance Strategies in Diffusion-Based Generative Models
Gyuseong Lee 고려대학교 대학원 2024 국내석사
In this study, we explore diffusion models, a type of generative model known for producing highly realistic images. Specifically, we investigate two guidance strategies—Classifier Guidance (CG) and Classifier-Free Guidance (CFG)—that are vital for enhancing the quality and control of images generated by these models. Our primary contribution is a new method for conditioning diffusion models to perform examplar-based image-to-image translation. We've incorporated techniques like matching and warping into the model's sampling process, allowing it to not only generate high-quality images but also adhere to specific conditions like edge maps. Another significant aspect of our work is addressing the limitations of CFG, particularly its dependence on external conditioning factors. We introduce Self-Attention Guidance (SAG), a novel guidance approach that not only boosts image quality without external conditioning but also improves results in combination with CFG. Our findings represent a major step forward in understanding and applying conditioning techniques in diffusion models, which could significantly enhance their ability to produce high-quality images and broaden their image generation capabilities.
연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
