作者：吴恩达

译自：Quora

找到新项目的灵感的一个好办法是花点时间研究以前的项目。

A great way for you to get ideas for new projects is to spend time studying previous projects.

大脑的工作就像某种魔法。当你研究了很多同类工作的例子（比如机器学习项目），你就会学会归纳总结，发明出这个类型的新例子。这也是为什么很多艺术家学习绘画的方法是复制前辈大师的作品——如果你去逛艺术馆，有时候你会看到一些艺术生坐在地板上试图复制墙上的艺术作品。类似的是，很多研究者用复制老论文结果的方法来发明新的算法。对我来说，我目睹了大量的业界不同公司的实际的机器学习用户案例，于是经常发现机器学习可以改造一些公司的新机会。

There’s something magical about how your brain works. When you study many examples of one category of work (such as ML projects), you will learn to generalize and invent new examples in that category. This is also why a lot of artists learn to paint by copying the work of the old masters—if you visit an art museum, sometimes you’ll sometimes see art students sitting on the floor trying to replicate the displayed artwork. Similarly, a lot of researchers learn to invent new algorithms by replicating the results in old research papers. For myself, it was through seeing a lot of actual ML use-cases in industry in different companies that’s helping me now regularly spot new opportunities for ML to transform companies.

所以，如果你想知道如何做有趣的项目，阅读（或者复制）你喜欢的以前的项目，然后你可以开始你自己的想法。作为一个起点，这里有一个项目是我的斯坦福学生最近做的：CS 229 机器学习最终项目，2016年秋季。

So, if you want to know how to do interesting projects, read (and perhaps replicate) previous projects that you liked, and you’ll start to get your own ideas. As a starting point, here’re projects that my Stanford students did recently: CS 229 Machine Learning Final Projects, Autumn 2016

最后当你完成了一个有趣的项目，请写一篇Arxiv论文或者blog，也许还可以在github开源你的代码，把它共享回社区。这样别人也可以从你这里学习。另外，你可能得到很多反馈，可以加速你的学习。

Finally, when you do complete an interesting project, please write an Arxiv paper or blog post, perhaps open-source your code on github, and share it back with the community! This way others can now in turn learn from you. Plus, you might also get more feedback, and thus accelerate your learning.

学习以前的例子以外，我还花时间跟人交谈——包括机器学习以外的专家（比如，我花了很多时间跟卫生保健专家聊）——这经常激发出机器学习和卫生保健或者其他领域交叉的新项目。

In addition to studying previous examples, I also spend time talking to people—including experts in areas other than ML (for example, I’ve spent a lot of time talking to healthcare experts)—and this often inspires new projects at the intersection of ML and healthcare or other areas.