Department Seminars & Colloquia
When you're logged in, you can subscribe seminars via e-mail
PET imaging can yield quantitative information about a radiotracer’s spatial and temporal distribution within the body. The ideal PET radiotracer will allow the detection of some changes at a very early stage of a disease or changes with treatment of that disease. In an ideal situation, the measure will be both quantitative and sensitive. However, in a clinical setting, it is less important for the tracer to be a quantitative measure than it is to be sensitive to the change. An arterial input function is typically measured by acquiring discrete arterial blood samples, usually from a radial artery. However the placement of the arterial catheter and frequent blood draws during the scan is also very difficult and is usually not performed in a clinical setting. These constraints limit the full quantification of the PET study. I will introduce the alternative to use an image-derived input function (IDIF) using the carotid artery. Also, I like to discuss how to quantify brain PET images with different input functions.
The problems that involve low rank constrained minimization of a given data matrix have attracted a great attention in recent years in data analysis including image analysis such as background modeling and face recognition. In this talk, we introduce a new formulation called linf-norm based nonnegative matrix factorization and its various properties, such as the relation between stability and sparsity of the proposed model. Numerical analysis shows positive performance of the proposed model compared to the state of the art model based no nuclear norm based rank minimization method.
주식/지수 파생상품의 이론가 산출에 쓰이는 변동성 데이터에 대해 소개한다. 특히 옵션의 시장 가격 데이터로부터 내재변동성 및 로컬 변동성 곡면을 산출해내는 방법을 단계별로 설명할 예정이다. 실제 시장 데이터에는 다양한 방식으로 노이즈가 개입될 수 있는데, 이런 노이즈 데이터를 적절히 필터링 해야 할 필요가 있다. 또한 필터링 된 후 남은 데이터가 변동성 곡면을 만들어 내기에 충분치 않을 수도 있다. 이와 같은 변동성 데이터 관련 이슈를 소개하고 그 해결책에 대해 논의한다.