가상계측 모델링을 위한 상관관계 정렬 전이학습

다국어 초록 (Multilingual Abstract)

The shrinkage of circuit line width, which is a key technology for semiconductor processes, has caused problems that small changes in process parameters result to affect in product quality. The metrology process is operated to figure out the quality problems in advance. However, virtual metrology study was conducted because monitoring of all products is impossible. In the virtual metrology study, the lack of learning data of new equipment, it is difficult to obtain an accurate prediction model. Therefore, transfer learning was also study to solve it. However, in case of transfer learning, the prediction accuracy is low when the existing equipment and the new equipment are not similar. In this paper, we propose a method of transfer learning with virtual metrology model by performing preprocessing using correlation alignment method for more accurate prediction of new equipment. The experimental study confirmed that the proposed approach shows superior performance in the results.

목차 (Table of Contents)

• 1. 서론 1
• 2. 제안 방법론 6
• 2.1. 도메인 적응: 상관관계 정렬 8
• 2.2. 전이학습: 미세조정 및 고정특징추출 11
• 3. 실험 및 결과 13
• 3.1. 평가 척도 13
• 3.2. 데이터 및 실험 설명 15
• 3.3. 데이터 평가 결과 18
• 4. 결론 27
• 참고 문헌 28