Approximate computing and scientific computing are two fields that seem incompatible at first glance. However, it turns out that these two fields not only complement each other, but can also improve and further develop each other. For example, suspended synchronization mechanisms can lead to more efficient solvers of linear systems of equations or methods such as automatic differentiation can provide starting points for approximation strategies.

Project start: 01.01.15
Project end:
Project group: Prof. Dr. Wolfang Karl, Markus Hoffmann
Contact:
Links: Jobs - Theses

The combination of approximate computing and scientific computing raises interesting questions for both computer science and mathematics. Can approximation strategies simply be applied to mathematical procedures? Can mathematical boundaries be overcome with approximate computing? How can suitable starting points for approximations be found within the procedures? Is the use of high precision also suitable for accelerating numerical processes? What role do schedulers, genetic algorithms, neural networks, provability and practical usability play in this environment? The tension between mathematical precision and approximate computing strategies offers a wide range of interesting research opportunities, both in computer science and mathematics.