The multi-scale, mutli-physics nature of fusion plasmas makes predicting plasma events challenging. Recent advances in deep convolutional neural network architectures (CNN) utilizing dilated convolutions enable accurate predictions on sequences which have long-range, multi-scale characteristics, such as the time-series generated by diagnostic instruments observing fusion plasmas.
Even though neural networks enjoy widespread use, they still struggle to learn the basic laws of physics. How might we endow them with better inductive biases? Recent work (Greydanus et al. 2019) proposed Hamiltonian Neural Networks (HNNs) which can learn the Hamiltonian of a physical system from data using a neural network. A key issue with these models is that they require a priori knowledge of the system’s conjugate position and momentum coordinates, and thus are difficult to learn from arbitrary coordinates such as pixels.
I will present a Bayesian machine learning architecture that combines a physically motivated parameterization and an analytic error model for the likelihood with a deep generative model providing a powerful data-driven prior for complex signals. This combination yields an interpretable and differentiable generative model, allows the incorporation of prior knowledge, and can be utilized for observations with different data quality without having to retrain the deep network.
Taking place every other Friday. Lunch will be provided.
This interdisciplinary meeting focusses on ML approaches that are useful for the sciences and engineering. The style is informal, a mix of journal club, practical tutorials, and project discussion from the participants. Join us if you want to learn new ML approaches for scientific research and, in particular, if you're already using them.
