PhD Defense: Disentanglement, Compositionality, Specification: Representation Learning with Generative Adversarial Networks

28 April 2021

Abstract: Learning good data representations is one of the most important tasks for machine learning. Generative adversarial nets (GANs) offer a powerful framework for learning image representations and offer functionalities such as image generation and editing. This thesis introduces three approaches to improve the learned image representations and their usage for several downstream tasks. Based on these approaches we show how we can use GANs to learn disentangled, compositional, and highly specific representations and how these can be applied to various tasks such as image generation, image-to-image translation, text-to-image synthesis, and image editing.

To participate in the PhD defense, please contact Tobias via email. The Zoom link will then be mailed to you shortly before the meeting.

Wednesday, 28 April 2021, 14:00, Online Meeting (zoom)

Speaker: Tobias Hinz, Informatics Dept., Univ. Hamburg