Mr. Sooho Kim has joined the Viper (Visual Information Processing for Enhanced Retreival) Group, co-organized by Université de Genève (UniGe) and Haute école de gestion de Genève (HES-SO) for visiting research as a visiting scholar. During visiting, he performs researches with two topics, which are analysis on latent feature space of Neural Networks and virtual data generation for PHM purpose.

The first topic, analysis on latent feature space of Neural Networks is an attempt to physically understand nonlinear and complex process of NN. Since PHM algorithm should suggest physical explanation on reasoning phase, this research will contribute to adopt NN approaches on PHM problem by interpreting physical background of NN process. The second topic, generation of virtual data is owing to enhancing PHM performance with limited amound of data. Based on physical properties from NN components, the research will extract and sort physical information from limitet amount of data, and will generate new virtual data by moderating and re-combining them.

The visiting period is March-June and September-December 2019. This research is supported by the Office of Interational Affairs (OIA), SNU.