Deep Learning Fellowship

The Deep Learning Fellowship is back for an in-person event. So, come and join a fellowship of deep learning and AI enthusiasts in their quest to keep up with the latest developments in artificial intelligence.
This group is targeted towards intermediate and advanced practitioners of deep learning, but all levels of expertise are welcome. The topics are advanced, but the mindset is that we are learning the material together.
As has been our tradition, we will select a paper ahead of time for each meeting from a vast pool of published research in the field of AI and Machine Learning. We will typically cover anything from computer vision, natural language processing to deep reinforcement learning and Bayesian networks. These meetups are intended to be an open discussion of the selected paper, related research and relevant model coding.
We are going to do something a bit different this time. We are going to cover the Dall-e (text-to-image) model and associated paper, but we are going to do it in three sessions. The Dall-e model has two major subcomponents: a discrete variational autoencoder (dVAE) and a transformer. So over three meetup sessions we will cover a transformer example, a dVAE example and then in the third session we will pull it all together to look at the Dall-e model.
The paper for this meetup is for the discrete variational autoencoder (dVAE):
"Categorical Reparameterization with Gumbel-Softmax" (https://arxiv.org/abs/1611.01144)
Some background on variational autoencoders can be found in the following:
"Tutorial on Variational Autoencoders" (https://arxiv.org/abs/1606.05908)
If you get a chance to read the paper before the meetup that is great. If not, don’t worry. We will be doing a detailed presentation covering the findings outlined in the paper, so if you don’t get a chance to read it you can still come to the session and join the discussion.
Our goal is to create an interactive study group where we collectively choose the research papers that we'd like to dig into and share the lessons we learn from attempting to execute the architecture in our own applications. Your feedback will be valuable in determining the structure and content of the event in the future.