Generative Modeling
STA 851
This course will focus on methodology for inferring generative probability models based on complex data. A focus will be on methods for density estimation, particularly focused on approaches that can be generalized to allow moderate to high-dimensional and complex data, and approaches that avoid explicit density estimation but can be used to generate future data from a distribution that is close to that of the training data. An additional focus will be on how to conduct interpretable inference. Many of the associated models can be viewed as variations on latent variable models; for example, based on starting with iid standard normal random variables and then applying nonlinear mappings. Approaches for inference will be described and illustrated through a variety of motivating applications.
Reserved for Statistical Science PhD students
Prerequisite: STA 702L and STA 831
3 Graduate Units
This course will focus on methodology for inferring generative probability models based on complex data. A focus will be on methods for density estimation, particularly focused on approaches that can be generalized to allow moderate to high-dimensional and complex data, and approaches that avoid explicit density estimation but can be used to generate future data from a distribution that is close to that of the training data. An additional focus will be on how to conduct interpretable inference. Many of the associated models can be viewed as variations on latent variable models; for example, based on starting with iid standard normal random variables and then applying nonlinear mappings. Approaches for inference will be described and illustrated through a variety of motivating applications.
Prerequisites
Prerequisite: STA 702L and STA 831