【深度学习】怎样在6个月内掌握深度学习

It is quite possible to learn,follow and contribute to state-of-art work in deep learning inabout 6 months’ time. This article details out the steps to achieve that.

利用大约6个月的时间，学习、跟随并在深度学习领域做出最新研究成果是完全有可能的。这篇文件详细介绍了实现这一目标的步骤。

Pre-requisites

- You are willing to spend 10–20 hours perweek for the next 6 months

- You have some programming skills. You should be comfortable to pick up Pythonalong the way. And cloud. (No background in Python and cloud assumed).

- Some math education in the past (algebra, geometry etc).

- Access to internet and computer.

先决条件

- 你愿意在接下来的6个月内每周花费10-20小时。

- 你有一些编程基础。你应该有一定的Python基础。有一些云计算基础。（本文假设你没有Python和云计算基础）。

- 有一定的数学基础（代数，几何等）。

- 熟练使用互联网和计算机。

Step 1

We learn driving a car — by driving. Not by

learning how the clutch and the internal combustion engine work. Atleast

not initially. When learning deep learning, we will follow the same top-down

approach.

Do the fast.ai course — Practical Deep Learning for Coders — Part 1. This takesabout 4–6 weeks of effort. This course has a session on running the code on cloud.Google Colaboratoryhas free GPU access. Start with that. Other optionsinclude Paperspace, AWS, GCP, Crestle and Floydhub. All of these aregreat. Do not start to build your own machine. Atleast not yet.

步骤1

我们学习汽车驾驶，不是从了解离合器和内燃机的工作原理开始。至少不是一开始就去了解这些。学习深度学习，我们同样也将遵循自上而下的方法。

学习fast.ai课程-Practical Deep Learning for Coders — Part 1。这需要花费大约4-6周的时间。这部分课程有如何在云上运行代码的介绍。Google Colaboratory拥有免费的GPU访问能力，可以从这里开始。其他的选择包括Paperspace,AWS,GCP,Crestle和Floydhub。所有这些都很棒。刚开始的时候，不要试图建立自己的环境。

Step 2

This is the time to know some of the

basics. Learn about calculus and linear algebra.

For calculus,Big Picture of Calculus provides a good overview.

For Linear Algebra, Gilbert Strang’s MITcourse on OpenCourseWare is amazing.

Once you finish the above two, readtheMatrix Calculus for Deep Learning.

第2步

是时候了解一些基础知识了。学习有关微积分和线性代数的知识。

对于微积分，Big Picture of Calculus提供了一个很好的概述。

对于线性代数，Gilbert Strang的MIT OpenCourseWare非常棒。

完成上述两项任务之后，请学习Matrix Calculus for Deep Learning。

Step 3

Now is the time to understand the bottom-upapproach to deep learning. Do all the 5 courses in the deep learning specialisation in Coursera. You need to pay to get theassignments graded. But the effort is truly worth it. Ideally, given thebackground you have gained so far, you should be able to complete one courseevery week.

第3步

现在是时候来理解深度学习自下而上的学习方法了。学习Coursera的deep learning specialisation的5门课程。你需要付费以便完成学习，但这些付出确实值得。理想情况下，根据你的背景，你应该可以每周完成一门课程。

Step 4

“All work and no play makes Jack a dullboy”

Do a capstone project. This is the time

where you delve deep into a deep learning library(eg: Tensorflow, PyTorch,

MXNet) and implement an architecture from scratch for a problem of your liking.

The first three steps are about

understanding how and where to use deep learning and gaining a solid

foundation. This step is all about implementing a project from scratch and

developing a strong foundation on the tools.

步骤4

“只工作不玩耍，聪明孩子也变傻”

去实践一个有难度的项目。现在是时候来深入研究一个深度学习库了（例如：Tensorflow，PyTorch，MXNet）并根据你的喜好选择一种架构来解决问题。

前三步是理解如何以及在哪里使用深度学习并为此打下坚实基础。这一步是开始实施项目，并为这些工具打下坚实的基础。

Step 5

Now go and do fast.ai’s part II course —Cutting Edge Deep Learning for Coders. This covers more advanced topics and you will learn toread the latest research papers and make sense out of them.

Each of the steps should take about 4–6 weeks’ time. And in about 26 weeks since the time you started, and if you followed all of the above religiously, you will have a solid foundation in deep learning.

第5步

现在去做fast.ai的第二部分课程-Cutting Edge Deep Learning for Coders。这涵盖了更高级的主题，你将学习研究最新的论文并从中理解其中的道理。

每个步骤大约需要4-6周的时间。如果你认真地完成上述所有学习内容，在你开始深度学习之后的第26周，你将在深度学习方面打下坚实的基础。

Where to go next?

Do the Stanford’s CS231n and CS224d courses.These two are amazing courses with great depth for vision and NLP respectively.They cover the latest state-of-art. And read the deep learning book.This will solidify your understanding.

Happy deep learning. Create every single day.

下一步该干什么？

做斯坦福的CS231n和CS224d课程。这两部分都是非常精彩的课程，分别对视觉和自然语言理解有很深的研究，它们涵盖了最新的技术。最后阅读deep learning book，这会进一步巩固你的理解。

Happy deep learning. Create every single day.

星星浩然，浩然之星！欢迎扫描二维码关注微信订阅号，或直接搜索“星星浩然”关注！

英文地址：https://towardsdatascience.com/how-to-learn-deep-learning-in-6-months-e45e40ef7d48