Teaching scientific computing through projects

Traditional numerical methods courses face several difficulties. There is a performance problem for students who need to combine different skills. Students spend so much time mastering the details that they often miss the point as to why there is a need for scientific computing. Using projects drawn from practical situations, and a good high-level package such as MATLAB, can overcome much of this. The difficulty associated with learning scientific programming is greatly alleviated. This in turn frees more time for studying more advanced methods. One needs to combine ideas from different areas makes this an ideal candidate for both team-teaching and student teamwork on the projects. In this paper, the authors discuss the use of this approach in a variety of scientific computing courses at different levels.