How to embed a data manifold by using locally linear maps?
陳志偉教授國立中山大學應用數學系 15:30 – 17:00 鴻經館 M107Abstract：
A data cloud with special pattern can be seen as a high dimensional (sub-)manifold and its features can be detected via various dimensionality reduction maps. Mechanisms of some well-known maps, such as LLE and HLLE, will be discussed in this talk. With Dr. Liren Lin, we propose a Tangential LLE which can capture intrinsic structure of the data manifold while LLE/HLLE fails or is non-applicable.