Code is often very important in current research, so it makes sense to include it. But code of any non trivial amount is difficult to prove correct. so it adds another level of complexity. And if you add data to drive the code, a kind of context, correctness can be further obscured. Still like the idea of including the code, to show an application of the research.
arXiv Now Allows Researchers to Submit Code with Their Manuscripts VentureBeat, By Khari Johnson
Machine learning resource Papers with Code said the preprint paper archive arXiv is now permitting researchers to submit code with their papers. A recent artificial intelligence (AI) industry report by London-based venture capital firm Air Street Capital found only 15% of submitted research manuscripts currently include code. Preprint repositories offer AI scientists the means to share their work immediately, before an often-protracted peer review process. Code shared on arXiv will be entered via Papers with Code, and the code submitted for each paper will appear on a new Code tab. Papers with Code co-creator Robert Stojnic said, "Having code on arXiv makes it much easier for researchers and practitioners to build on the latest machine learning research. We also hope this change has ripple effects on broader computational science beyond machine learning."
No comments:
Post a Comment