Guest Lectures in coming Interactive Sessions

Hi everyone!

In this and the next 2 weeks, we will have visits from experts on topics related to the weekly syllabus in the interactive session. The first one, on Thursday April 4th, is postdoctoral researcher Benedikte Wallace, who will talk about how to apply unsupervised / self-supervised learning for modelling dance movements. See some more info below - hope to see many of you there!

-Kai

--------------

Benedikte will talk about her work on using deep learning to generate dance movement using a dataset of motion capture data. Her work uses sequence prediction models such as the Transformer and the RNN to predict the next position of the dancer¡¯s body. While these models learn the most likely positions from the data, there are many poses that might be a good fit!

How do we decide what makes a good dance move? What makes something interesting or beautiful? In this lecture we will look at some of these challenges and the methods that are commonly used for generating creative data.