Facilitating Reproducibility in Scientific Computing: Principles and Practice

Abstract

The foundation of scientific research is theory and experiment, carefully documented in open publications, in part so that other researchers can reproduce and validate the claimed findings. Unfortunately, the field of scientific and mathematical computing has evolved in ways that often do not meet these high standards. In published computational work, frequently, there is no record of the workflow process that produced the published computational results, and in some cases, even the code is missing or has been changed significantly since the study was completed. In other cases, the computation is subject to statistical errors or numerical variability that makes it difficult for other researchers to reconstruct results. Thus confusion often reigns. That tide may be changing, though, in the wake of recent efforts that recognize both the need for explicit and widely implemented standards and also the opportunity to do computational research work more effectively. This chapter discusses the roots of the reproducibility problem in scientific computing and summarizes some possible solutions that have been suggested in the community.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 21 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.