CSE 708 / Symbolic Computing and Automatic Differentiation

1.5 unit(s)

Staff

This module presents an overview of techniques for symbolic computations and in particular for computing derivatives arising in scientific computing applications. Then it introduces the building blocks of automatic differentiation (AD) and how AD is implemented in software tools through source-code translation and operator overloading. In addition to computing first-order derivatives, students will learn how second and higher-order derivatives are computed through AD, and techniques for exploiting sparsity. Applications to optimization, sensitivity analysis, and machine learning will be considered.