Advancing Space Science with Machine Learning: Frontier Development Lab Projects with NASA

We'll briefly introduce the NVIDIA-supported Frontier Development Lab, a partnership between the SETI Institute and NASA, then present results from three teams: Astrobiology (detecting atmospheric molecules in exoplanetary atmospheres), Astronaut Health I (generative models for ECG astronaut health data generation), and Astronaut Health II (a causal inference platform for modeling cancer progression). Each team faced world-class challenges and used machine learning to make significant advances in their respective domains. We'll outline each challenge and describe each team's high-impact solutions. Machine learning methods include variational autoencoders, Bayesian neural networks, generative models, and the latest causal inference methods. Computational aspects range from single GPU usage and dedicated IBM systems to massive CPU and GPU utilization on Google Cloud (V100 GPUs) to generate (3M) entire exoplanetary atmospheres.