Advanced Topics in AI and Robotics

Abstract

In her lecture, Shinde gives an introduction to the self-attention mechanism, a very simple and powerful sequence-to-sequence layer that is at the heart of transformer architectures. The lecture explains the basic idea behind self-attention, why it works so well, and its practical implementation. In the end, the application of this mechanism for estimating the pose of the robot using sensor fusion is demonstrated.

Short Bio

Kashmira Shinde is a researcher in the field of Robotics and Deep Learning at the Robotics Group of the University of Bremen. She is currently working on developing a flexible software framework for human-robot interaction in space scenarios. She received a Master of Science degree in Automation and Robotics with a specialization in Robotics from the Technical University of Dortmund in 2020 and obtained her Bachelor's degree in Electronics and Instrumentation Engineering from India in 2016. Her interests lie in devising new robot perception and localization approaches by leveraging AI.

Papers relevant for the talk

1. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.

2. Bloem, P. (2019, August 19). Transformers from scratch. https://peterbloem.nl/blog/transformers

3. Shinde, K., Lee, J., Humt, M., Sezgin, A., & Triebel, R. (2020). Learning multiplicative interactions with bayesian neural networks for visual-inertial odometry. arXiv preprint arXiv:2007.07630

The lecture will be held in English and is aimed at students and staff of the H-BRS. Interested parties are cordially invited.