Abstract:Differentiable programming allows for derivatives of functions implemented via computer code to be calculated automatically. These derivatives are calculated using automatic differentiation (AD). This thesis explores two applications of differentiable programming to computational plasma physics. First, we consider how differentiable programming can be used to simplify and improve stellarator optimization. We introduce a stellarator coil design code (FOCUSADD) that uses gradient-based optimization to produce stellarator coils with finite build. Because we use reverse mode AD, which can compute gradients of scalar functions with the same computational complexity as the function, FOCUSADD is simple, flexible, and efficient. We then discuss two additional applications of AD in stellarator optimization. Second, we explore how machine learning (ML) can be used to improve or replace the numerical methods used to solve partial differential equations (PDEs), focusing on time-dependent PDEs in fluid mechanics relevant to plasma physics. Differentiable programming allows neural networks and other techniques from ML to be embedded within numerical methods. This is a promising, but relatively new, research area. We focus on two basic questions. First, can we design ML-based PDE solvers that have the same guarantees of conservation, stability, and positivity that standard numerical methods do? The answer is yes; we introduce error-correcting algorithms that preserve invariants of time-dependent PDEs. Second, which types of ML-based solvers work best at solving PDEs? We perform a systematic review of the scientific literature on solving PDEs with ML. Unfortunately we discover two issues, weak baselines and reporting biases, that affect the interpretation reproducibility of a significant majority of published research. We conclude that using ML to solve PDEs is not as promising as we initially believed.

Experts' understanding of partial derivatives using the Partial Derivative Machine

Promising directions of machine learning for partial differential equations

What You See is Not What You Get: Neural Partial Differential Equations and The Illusion of Learning

The development of students' graphical understanding of the derivative

Machine Learning for Partial Differential Equations

Promoting Prospective Mathematics Teachers' Understanding of Derivative across Different Real-life Contexts

An energy approach to the solution of partial differential equations in computational mechanics via machine learning: Concepts, implementation and applications

Experts’ Construction of Mathematical Meaning for Derivatives and Integrals of Complex-Valued Functions

Towards understanding and characterizing expert covariational reasoning in physics

An Overview on Machine Learning Methods for Partial Differential Equations: from Physics Informed Neural Networks to Deep Operator Learning

Deep learning methods for partial differential equations and related parameter identification problems

Nonparametric estimation via partial derivatives

A two-phase study examining perspectives and use of quantitative methods in PER

Exactness: A Concept Important for Engineering Applications or a Source of Potential Difficulties?

Perspectives of physics-based machine learning strategies for geoscientific applications governed by partial differential equations

Students' Experience of Cultural Differences Between Mathematics and Physics

Partial Differential Equations Meet Deep Neural Networks: A Survey

Differentiable Programming for Computational Plasma Physics

Assessing Student Expertise in Introductory Physics with Isomorphic Problems, Part I: Performance on Non-intuitive Problem Pair from Introductory Physics

Comparing AI and student responses on variations of questions through the lens of sensemaking and mechanistic reasoning