TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Supporting the ML Reproducibility Challenge

1 点作者 dpleban大约 4 年前
Hey, HN! I&#x27;m one of the creators of DAGsHub(https:&#x2F;&#x2F;dagshub.com). As some of you might know, there&#x27;s an ML Reproducibility Challenge(https:&#x2F;&#x2F;paperswithcode.com&#x2F;rc2020) happening now, organized by Papers with Code, with some of the leading ML conferences. If you haven&#x27;t heard about it, or didn&#x27;t consider joining, I think that it provides an awesome opportunity for a number of reasons:<p>- It&#x27;s a chance to work on SOTA research, with a clearly defined goal – I personally took part in the last round of the challenge and it was a lot of fun and very challenging (I learned a lot from the other participants). - You can create a cool project for your portfolio. - You&#x27;ll be helping to advance the field by making sure that research results are replicable.<p>I personally care a lot about ML reproducibility, and making machine learning projects easily reproducible is one of the reasons we started DAGsHub in the first place.<p>We&#x27;ve decided to support the challenge, by providing participants $500 per paper reproduced, to help cover compute costs and incentivize participation, as well as provide our platform and support from our team and community for the challenge.<p>This is not a challenge we started, and other organizations are supporting it as well, but we thought this might encourage more people to take part, and think that everyone would benefit.<p>You can read the full guidelines here: https:&#x2F;&#x2F;dagshub.com&#x2F;pages&#x2F;reproducibility-challenge<p>Would love to answer any questions you might have.

1 comment

yesenadam大约 4 年前
If you write a blog post saying this and submit it as a usual HN story, or submit an existing webpage as a story, a lot more people might see it. I think many people never look in AskHN, where I found this (like everything submitted without URL), and it doesn&#x27;t really fit. Good luck!