Time: | May 29, 2024, 2:00 p.m. (CEST) |
---|---|
Venue: | 8.122 Pfaffenwaldring 57 70569 Stuttgart |
Download as iCal: |
|
We are happy to announce the next presentation in the ML-Session series:
Alina Roitberg (University of Stuttgart) will present on Wednesday 29 May 2024 at 2pm in PWR 57, 8.122 a in person lecture on "Foundation Models"
Abstract: Foundation Models, such as GPT4, Stable Diffusion and CLIP, are neural networks trained on very large amounts of data, which can be adapted for various downstream tasks with minimal or no additional data. This lecture explores the concept and applications of Multimodal Foundation Models (FMs). We will start with a historical overview of the field, leading to the emergence of FMs and discuss their definition. Special emphasis will be placed on Vision and Vision-and-Language Models, briefly touching upon Language-only FMs and concluding with recent advances in FMs for additional modalities. The session includes relevant building blocks and optimization techniques such as transformer models, self-supervised and contrastive learning, alongside examples of recent architectures developed for specific tasks. We will conclude with insights into FMs for text-guided image synthesis and data-scarce modalities beyond vision and language.