.png?sfvrsn=d3a97fa3_12)
Differentiable Physics Simulations for Deep Learning
[CFAR Rising Star Lecture Series]
Differentiable Physics Simulations for Deep Learning (Hybrid Event) by Prof Nils Thuerey
This talk focuses on the possibilities that arise from recent advances in deep learning for physical simulations. It would focus on differentiable physics solvers from the larger field of differentiable programming. These solvers provide crucial information for deep learning tasks in the form of gradients, which are especially important for time-dependent processes. Similarly, the existing numerical methods for efficient solvers could be leveraged within learning tasks. This paves the way for hybrid solvers in which traditional methods work alongside with pre-trained neural network components.
The resulting improvements would be illustrated with examples such as wake flows and turbulence mixing layer cases. From a machine learning perspective, regression problems with physics solvers are a highly interesting class of problems. Prof Nils Thuerey would conclude the talk by outlining avenues for custom learning algorithms that leverage the information from the solvers at training time.
Differentiable Physics Simulations for Deep Learning (Hybrid Event) by Prof Nils Thuerey
12 Jan 2023 | 2.00pm (òòò½ÍøTime)
This talk focuses on the possibilities that arise from recent advances in deep learning for physical simulations. It would focus on differentiable physics solvers from the larger field of differentiable programming. These solvers provide crucial information for deep learning tasks in the form of gradients, which are especially important for time-dependent processes. Similarly, the existing numerical methods for efficient solvers could be leveraged within learning tasks. This paves the way for hybrid solvers in which traditional methods work alongside with pre-trained neural network components.
The resulting improvements would be illustrated with examples such as wake flows and turbulence mixing layer cases. From a machine learning perspective, regression problems with physics solvers are a highly interesting class of problems. Prof Nils Thuerey would conclude the talk by outlining avenues for custom learning algorithms that leverage the information from the solvers at training time.
SPEAKER
Prof Nils Thuerey
Associate-Professor
Technical University of Munich (TUM)
Associate-Professor
Technical University of Munich (TUM)
Prof Nils Thuerey is an Associate-Professor at the Technical University of Munich (TUM). He focuses on deep-learning methods for physical systems, with an emphasis on fluid flow problems. Together with his research group, he is especially interested in methods for the tight and seamless integration of learning algorithms with numerical methods, i.e., via differentiable physics simulations. Additionally, Prof Thuerey also works on latent-space simulation algorithms, generative adversarial models, and visual reconstructions.
A*STAR celebrates International Women's Day
From groundbreaking discoveries to cutting-edge research, our researchers are empowering the next generation of female science, technology, engineering and mathematics (STEM) leaders.