David Defour
Some issues with Floating-Point representation formats and their numerical meaning
In the realm of computational science, numerical accuracy and reproducibility are critical to ensuring reliable results. This talk delves into the intricacies of floating-point representation, emphasizing the challenges posed by the large number of extremely small numerical formats and the effect of rounding errors, processor, compiler, operating system and programming language.
The impact of these limitations on reproducibility, accuracy, and repeatability of scientific computations and IA will be a focal point. Additionally, we will examine the implications for the meaningful interpretation of results when using ultra-small numerical representations. This talk aims to provide a comprehensive understanding of the balance between computational efficiency and numerical fidelity, offering insights, best practices and link for researchers and practitioners working with constrained numerical formats.
The impact of these limitations on reproducibility, accuracy, and repeatability of scientific computations and IA will be a focal point. Additionally, we will examine the implications for the meaningful interpretation of results when using ultra-small numerical representations. This talk aims to provide a comprehensive understanding of the balance between computational efficiency and numerical fidelity, offering insights, best practices and link for researchers and practitioners working with constrained numerical formats.
back to overview
