GTC 2020: Inductive Biases for Robot Learning
After clicking “Watch Now” you will be prompted to login or join.
Click “Watch Now” to login or join the NVIDIA Developer Program.
Inductive Biases for Robot Learning
Michael Lutter, TU Darmstadt
While deep learning has propelled many applications, including computer vision and natural language processing, it hasn't enabled physical robots to achieve new tasks. We'll introduce the challenges of working with robots and the possibilities of deep learning for robotics, and we'll highlight the importance of constraining deep networks with domain knowledge to enable learning of controllers and models suitable for the physical world. Besides the application to robotics, our concepts for combining domain knowledge and deep learning also apply to other engineering disciplines and the natural sciences.