ai 人工智能 入门
介绍 (Introduction)
Focus on research in Artificial Intelligence (AI) is nowadays growing more and more every year, particularly in fields such as Deep Learning, Reinforcement Learning and Natural Language Processing (Figure 1).
如今，对人工智能(AI)研究的关注每年都在增长，尤其是在深度学习，强化学习和自然语言处理等领域(图1)。
State of the art research in AI is usually carried out in top universities research groups and research-focused companies such as Deep Mind or Open AI, but what if you would like to give your own contribution in your spare time?
人工智能的最先进研究通常在顶级大学研究小组和专注于研究的公司(例如Deep Mind或Open AI)中进行，但是如果您想在业余时间做出自己的贡献，该怎么办？
In this article, we are going to explore different possible approaches you can take in order to be always up to date with the latest in research and how to provide your own contribution.
在本文中，我们将探索您可以采取的各种可能的方法，以便始终与最新的研究保持同步，并提供自己的贡献。
再现性挑战 (The Reproducibility Challenge)
One of the main problems which have affected the AI research field is the possible inability to efficiently reproduce models and results claimed in some publications (Reproducibility Challenge).
影响AI研究领域的主要问题之一是可能无法有效地重现某些出版物中提出的模型和结果(可再现性挑战)。
In fact, many research articles published every year contains just an explanation of the derided topic and model developed but no source code to reproduce their results. Some reasons why researchers might at times omit these kinds of information are: keep a competitive advantage against other institutions, non-disclosure agreements, transform their research into a product, etc…
实际上，每年发表的许多研究文章仅包含对所开发的主题和模型的解释，而没有可重现其结果的源代码。 研究人员有时可能会忽略此类信息的一些原因是：与其他机构保持竞争优势，保密协议，将研究成果转化为产品等…
In order to make research more accessible and have real-world impacts, different competitions have been created in order to encourage the public to study different publications and try to reproduce their results. Two of the most know competitions in this ambit are the NeurIPS and ICLR Reproducibility Challenges. In case you are looking for any practical example, I recently started a GitHub repository about this topic.
为了使研究更容易获得并具有现实世界的影响，为了鼓励公众学习不同的出版物并尝试重现其成果，开展了各种竞赛。 在此领域中，最知名的两个竞赛是NeurIPS和ICLR再现性挑战。 如果您正在寻找任何实际的例子，我最近就此主题启动了一个GitHub存储库 。
Additionally, websites like Papers with Code, have recently been created in order to easily find research publications which already have publicly available code. In this way, anyone can use state of the art models for their own projects completely for free!
此外，最近还创建了带有代码的论文之类的网站，以便轻松查找已经具有公开代码的研究出版物。 这样，任何人都可以完全免费地为自己的项目使用最新模型！
文件季节 (Season of Docs)
Season of Docs is an annual program organised by Google aimed at connecting technical writers with open-source organizations in order to improve libraries documentation.
Google文件季节是Google举办的一项年度计划，旨在将技术作家与开放源代码组织联系起来，以改善图书馆的文档编制。
By joining the program, writers will, in fact, be able to contribute to the documentation of open-source organizations such as Julia, Numpy, Matplotlib, Bokeh and many more.
通过加入该计划，作者实际上将能够为开源组织的文档做出贡献，例如Julia，Numpy，Matplotlib，Bokeh等。
GitHub开源贡献 (GitHub Open Source Contributions)
Many of nowadays most popular Machine Learning and Deep Learning libraries are available on GitHub and most of them are happy to accept help from external contributors. Some examples of popular GitHub repositories with many Issues and Pull Requests which accepts contributors are:
如今，许多最受欢迎的机器学习和深度学习库都可以在GitHub上获得，其中大多数都乐于接受外部贡献者的帮助。 受欢迎的GitHub存储库中有许多Issues和Pull Requests可以接受贡献者的示例如下：
PyTorch
火炬
TensorFlow
TensorFlow
Keras
凯拉斯
Scikit-learn
Scikit学习
PySyft
PySyft
In case you are looking to explore for more of the available project, GitHub Collections are a great place from where to start (eg. Machine Learning).
如果您正在寻找更多可用项目，则GitHub Collections是一个很好的起点(例如， Machine Learning )。
两分钟论文 (Two Minute Papers)
Another way in order to keep always up to date with the latest research is to follow online publications like Towards Data Science and research focused YouTube channels like Two Minute Papers.
与时俱进的另一种方法是关注诸如Towards Data Science之类的在线出版物和关注诸如两分钟论文之类的YouTube重点频道。
This YouTube channel in fact reviews and summarises for you on a weekly bases some of the most interesting latest publications, providing also demos and example applications.
实际上，这个YouTube频道每周都会为您回顾和总结一些最有趣的最新出版物，还提供了演示和示例应用程序。
附加功能 (Extras)
Finally, other possible ways in order to keep always updated about AI is to:
最后，为了使AI始终保持最新状态的其他可能方法是：
Follow important personalities in the field such as Cassie Kozyrkov, Andrej Karpathy and Andrew Ng.
关注该领域的重要人物，例如Cassie Kozyrkov ， Andrej Karpathy和Andrew Ng 。
Take part in conference events such as: NeurIPS (Neural Information Processing Systems), ICLR (International Conference on Learning Representations), ICML (International Conference on Machine Learning) and AAAI (Association for the Advancement of Artificial Intelligence), etc….
参加会议活动，例如： NeurIPS (神经信息处理系统)， ICLR (国际学习代表会议)， ICML (国际机器学习会议)和AAAI (人工智能促进协会)等。
Read curated journals such as Distill, Fermat’s Library and Papers We Love.
阅读精选的期刊，例如Distill ， Fermat的图书馆和We Love Papers 。
If you have any suggestion on any other possible technique which can be added to this list, please just let me know in the comment section!
如果您对可以添加到此列表的任何其他可能的技术有任何建议，请在评论部分中告诉我！
I hope you enjoyed this article, thank you for reading!
希望您喜欢这篇文章，感谢您的阅读！
联络人 (Contacts)
If you want to keep updated with my latest articles and projects follow me on Medium and subscribe to my mailing list. These are some of my contacts details:
如果您想随时了解我的最新文章和项目，请在Medium上关注我，并订阅我的邮件列表 。 这些是我的一些联系方式：
Linkedin
领英
Personal Blog
个人博客
Personal Website
个人网站
Patreon
Patreon
Medium Profile
中档
GitHub
的GitHub
Kaggle
卡格勒
参考书目 (Bibliography)
[1] Artificial Intelligence Index 2018 Annual Report by Yoav Shoham et. al. Accessed at: http://cdn.aiindex.org/2018/AI%20Index%202018%20Annual%20Report.pdf
[1]人工智能指数2018年年度报告，作者Yoav Shoham等。 等 访问： http : //cdn.aiindex.org/2018/AI%20Index%202018%20Annual%20Report.pdf
[2] Reactome, Season of Docs. Accessed at: https://reactome.org/about/news/136-season-of-docs
[2] Reactome，文件季。 访问网址 ： https : //reactome.org/about/news/136-season-of-docs
翻译自: https://towardsdatascience.com/getting-started-in-ai-research-a683e6845537
ai 人工智能 入门