Deep Generative Models

This lecture covers the fundamentals of deep generative models, starting with a recap on document analysis and mixture of multinoullis. It then delves into Latent Dirichlet Allocation (LDA) and its generative model, discussing plate diagrams and learning methods. The lecture further explores variational inference for LDA, mean-field variational inference, and the training of autoencoders. It concludes with an introduction to variational autoencoders, their training process, and the challenges faced in training Generative Adversarial Networks (GANs). The lecture also touches on Deep Convolutional GANs (DCGANs) and other generative models like DALL-E 2.