Complex Numerical Analysis
Summary
This group activity is designed for experimenting and discovering the advantages of Complex Numerical Analysis. The submitted module is particularly for numerical differentiation. Through this Group Work, students may understand the limitations of finite precision numeric computation (machines round off errors, especially due to the "subtractive cancellation" in finite precision arithmetic), and may discover an advantage of using complex variables in numerical algorithms. If any student shows interest in this topic, the instructor may lead further development of the topic as an independent study or research for other numerical algorithms.
Learning Goals
Students will learn about issues in finite difference schemes in finite precision arithmetic, and how to avoid the problem. That is,
- Understand the limitation of finite precision arithmetic, especially the subtractive cancellation
- Discovery of advantage by using complex variables (in step size) in numerical differentiation
In addition, one may also address
- Order of convergence (of a numerical method)
Context for Use
This module is a part of the "Experimental Mathematics" class group activities. In this class, we performed a series of computational experiments and discoveries through Group Work in a computer lab setting. However, this module can be adopted in any Numerical Analysis or Scientific Computation class, as it addresses some basic but important issues in finite precision arithmetic and an idea of how to resolve the issues by introducing complex variables. This module can easily be extended to an undergraduate research topic.
MATLAB is mainly used and illustrated as an experimental tool to discover the contents. Students will experience both vibe-coding and peer-coding to generate and understand the needed code for experiments.
Description and Teaching Materials
We attached the Group Work (Complex Numerical Analysis I), a note for instructors (for the Group Work), two MATLAB sample codes and a key reference.
Student Handout for "Complex Numerical Analysis I" (Acrobat (PDF) 30kB Sep30 26)
Forward Finite Difference Scheme (Matlab File 2kB Sep30 26)
Complex Step Differentiation (Matlab File 1kB Nov8 26)
Numerical Differentiation of Analytic Functions (Acrobat (PDF) 499kB Sep30 26)
Teaching Notes and Tips
Teaching Note is posted in the file "Instructor Note.pdf"
Assessment
Each group is supposed to submit a group report for the class activities. Our class adopted mastery-based grading. If the class schedule allows, students can also present their learning in class.
References and Resources
- Numerical Differentiation - Wikipedia
https://en.wikipedia.org/wiki/Numerical_differentiation - Lyness, James N., and Moler, Cleve B., Numerical Differentiation of Analytic Functions, SIAM J. Numerical Analysis 4, 1967, pp. 202-210. https://epubs.siam.org/doi/abs/10.1137/0704019
- Squire, William, and Trapp, George, Using complex variables to estimate derivatives of real functions, SIAM Review 40, 1998, pp. 110-112.
https://epubs.siam.org/doi/abs/10.1137/S003614459631241X - Cleve Moler, Complex Step Differentiation, 2013
https://blogs.mathworks.com/cleve/2013/10/14/complex-step-differentiation/ - Most Numerical Method/Analysis textbooks have a topic on numerical differentiation (finite difference method) and its analysis.