What can auto-differentiation do for you? Applications of machine-learning tools to stars, dynamics and dust
Abstract
Machine learning has found increasing use in astronomy over the last few years. In this talk, instead of discussing machine learning itself, I will discuss how a mathematical tool borrowed from machine learning – auto-differentiation – can be put to use in astronomy. I will focus on applications of auto-differentiation to modeling stellar spectral energy distributions, gravitational potentials and the three-dimensional distribution of interstellar dust. The ability to construct highly flexible yet differentiable models allows elegant and straightforward approaches to solving each of these three problems.