机器学习实验室博士生系列论坛（第十五期）—— Recent Progress on Generative Modeling
报告人：Weijian Luo (PKU)
时间：2021-09-29 15:10-16:10
地点：北大理科一号楼1513会议室 & 腾讯会议 761 4699 1810
Abstract: Generative modeling has long been studied in machine learning field. Certain generative models such as generative adversarial network and variational auto-encoder has numerous application in representation learning, data augmentation, molecule generation, et al. However two special new forms of generative models has been intensely studied especially in recent years. One is normalizing flow which is based on the idea of transforming a simple distribution into data distribution while the other is SDE based generative modeling methods which use property of diffusion process to define the change of distribution. Recent work find out SDE based generative models can give best interpretable performance on generative tasks. Some recent work on Normalizing Flow and SDE based generative models will be given in this talk. We will especially focus on general methods and principles behind them.
