Published in the Journal of Computational Science, this paper shows how SmartSim was used to augment an ocean model (MOM6) with an AI-based, eddy kinetic energy parameterization to improve the climatology of the model. This particular implementation improved the RMSE of the objective metric by 20% compared to the reference simulation.
Pangeo Showcase talk about using SmartSim for online analytics, processing, and inference for HPC simulations. Pangeo is awesome. Great group of folks.
This paper explores Generative Teaching Networks (GTN), which are similar to GANs but instead of compete, two networks cooperate on a task. With applications in multiple domains, GTNs can aid supervised learning training times, learn Reinforcement Learning tasks like cart-pole, and perform Neural Architecture Search (NAS).
Threading in Python can sometimes be very confusing and frustrating. Understanding when to use the Threading module can make things much easier. These notes give some light background and some examples of good threading usage.