The full text 3412 words, when learning is expected to grow 10 Fenzhong
Source: Pexels
I was in mechanical engineering, the university became a mechanical engineer. My career began in a core work in the steel industry.
Wearing those heavy steel rubber boots and plastic helmets, work in large-scale blast furnace and rolling mill in. These security measures are just some psychological comfort, because I know that if something bad happens, they are not used. Perhaps running shoes can help. As for the helmet, I just want to say molten steel at 1370 degrees Celsius will be reduced to ashes.
With my deep fear of getting the job, I realized that this job is not for me, so I set a goal, probably into the data analysis and scientific fields in 2011. Since then, Mu class has become the preferred platform for me to learn new knowledge, through which I gain a lot of new knowledge. There are good and bad there.
Today, 2020, the ever-changing field of scientific data, the scientific data is not a lack of learning resources. But it also often cause problems for beginners: where to start learning? What to learn? There are many excellent resources on the Internet, but at the same time there are a lot of good resources.
Too much choice but to let people stagnate, because anxiety is to learn as the enemy.
Schwartz (Schwartz) in his book "The Paradox of Choice - Why More is Less but" in the proposed reduction of consumer choice can greatly reduce their anxiety. Data science curriculum as well.
This article provides suggestions for learners who lost the recommended starting point for some of the data science trip.
1) Python3 specialized programming
Python2.7的“Goodbye World”！
First, you need to determine a programming language. This is a specialized course at the University of Michigan, you can learn to use Python and create their own new things.
You will learn the basics of programming variables, conditionals and loops, etc., and get some intermediate materials, such as keyword arguments, list comprehension, lambda expressions, and class inheritance.
2) Python application data science
Do first, after understanding
Before fully understanding machine learning, we need to experience it.
Python application data science to introduce you to a lot of modern machine learning methods should be aware of. Not complete the run, but you'll get the tools you build the model.
This has the basic python programming background or skill-based professional-oriented and want to make a python by popular toolkits (such as pandas, matplotlib, scikitlearn, nltk and networkx) application statistics, machine learning, information visualization, text analysis and social networks learner analysis techniques to gain insight into their data.
3) machine learning theory and foundation
Source: Pexels
After completing the course, you will become a so-called "beginner."
Congratulations! ! ! You already know some basic concepts, but also know how to implement some of the features of the.
You are valuable
However, you do not fully understand all of these models and mathematical principles behind the model.
You need to understand the logic behind clf.fit. You have to face reality: Unless you understand the mathematical principles behind the model, otherwise nobody would really recognize you.
If you do not understand, you can not improve it
GameChanger machine learning curriculum includes many mathematical logic behind the machine learning algorithms.
I will pack this course as a required course, this course because it inspired me to enter this field, and Andrew Ng is a great teacher, this is my first learning courses.
This course includes regression, classification, anomaly detection, recommendation systems, neural networks, there are a lot of great advice.
4) study of statistical inference
"The truth is fixed, but the statistic is flexible." - Mark Twain
My mentor Chedi Kaja Lendl (Çetinkaya-Rundel) teaching this inferential statistics, it is easy to learn.
She was an excellent instructor, very well explain the basic principles of statistical inference - which is a required course.
You will learn hypothesis testing, confidence intervals, and statistical inference methods of numerical and categorical data.
SQL basics 5) learning scientific data
SQL is the core of all the data ETL
Although we feel that by creating a model and made different assumptions can accomplish more work, but the data chew role can not be underestimated.
And with the wide range of applications in SQL ETL and data preparation tasks, everyone should understand them slightly, because it is at least useful.
SQL has also become a big data tools such as the use of ApacheSpark facto standard. This UCDavis of SQL specialization course will teach you how to use SQL and SQL distributed computing.
通过使用数据科学应用程序的四个逐步增加难度的SQL项目，你将了解诸如SQL基础知识、数据争用、SQL分析、AB测试、使用ApacheSpark的分布式计算等主题。
6) 高级机器学习
在大联盟里，没有填鸭式灌输。
你可能不同意，但到目前为止，我们所做的一切都是骗人的。材料有固定的结构的，数学原理很少说明。但你已经为下一步做好了准备。这种高级机器学习专业化由顶级Kaggle机器学习实践者和欧洲核子研究中心的科学家采用了另一种学习方法，通过经历许多困难的概念，并指导您了解过去的事情是如何工作的，以及机器学习世界中最新的进步。网站上的描述是：
该专业介绍了深度学习、强化学习、自然语言理解、计算机视觉和贝叶斯方法。顶级Kaggle机器学习实践者和CERN科学家将分享他们解决现实世界问题的经验，并帮助你填补理论和实践之间的空白。
7) 深度学习
来源：Pexels
深度学习乃未来大势所趋
吴恩达带着全新的深度学习专业再次回归。
他以一种通俗易懂的方式完成了对这个困难概念的理解。他所遵循的术语与网络上所有其他教程和课程都不一样，我希望它能流行起来，因为这对理解所有基本概念非常有帮助。
专业网站上有说：
了解深层学习的基础，了解如何构建神经网络，并学习如何领导成功的机器学习项目。你将了解卷积网络、RNNs、LSTM、Adam、Dropout、BatchNorm、Xavier/He初始化等。且能接触医疗保健、自动驾驶、手语阅读、音乐生成和自然语言处理等方面的案例研究。
8) Pytorch
PythononFire
我通常从不提倡学习工具，但这不一样，因为如果你了解Pythorch，你便能够在许多最近的研究论文中学习代码，这真的很难得。Pythorch已经成为从事深度学习的研究人员的默认编程语言，它会对我们的学习极有帮助。
学习Pythorch的一种结构化方法是使用Pythorch学习深神经网络课程。课程网站说明：
课程将从Pytorch的张量和自动微分包开始。然后每个部分将涵盖不同的模型，从基本原理开始，如线性回归和logistic/softmax回归。其次是前馈型深层神经网络，作用不同的激活函数，归一化层和脱落层。然后介绍卷积神经网络和转移学习。最后，还将介绍其他一些深度学习方法。
9) AWS机器学习入门
秘诀：不是你知道什么，而是你展示什么。
在构建一个伟大的机器学习系统时，有很多事情需要考虑。但作为数据科学家，我们常常只担心项目的某些部分。
但我们有没有想过，一旦我们拥有了模型，要如何部署它们呢？
我见过很多ML项目，其中很多项目注定要失败，因为它们从一开始就没有一个固定的生产计划。
拥有一个好的平台，并了解该平台如何部署机器学习应用程序，将在现实世界中发挥所有作用。这门关于实现机器学习应用程序的AWS的课程就承诺了这一点。
本课程将教你：
1.如何使用带有内置算法和Jupyter笔记本实例的AmazonSageMaker构建、培训和部署模型。
2．如何使用Amazon-AI服务构建智能应用程序，如Amazon-Comprehend、Amazon-Rekognition、Amazon-Translate等。
10) 数据结构和算法
来源：Pexels
算法。是的，你需要它们。
算法和数据结构是数据科学的组成部分。虽然大多数的数据科学家在学习的时候并没有学习一门正确的算法课程，但它们是必不可少的。
许多公司要求将数据结构和算法作为招聘数据科学家面试内容的一部分。
它们需要和你的数据科学访谈一样的热情去破解，因此，你可能需要一些时间来研究算法、数据结构和算法问题。
我认为学习算法的一个最佳资源是UCSanDiego在Coursera上的算法专项课程。专业网站显示：
你将学习解决各种计算问题的算法技术，并将使用选择的编程语言实现大约100个算法编码问题。没有任何一门其他的在线算法课程能让你在下一次面试中有如此丰富的面临编程挑战经验。
希望你有所收获~
留言 点赞 关注
我们一起分享AI学习与发展的干货
欢迎关注全平台AI垂类自媒体 “读芯术”
（添加小编微信：dxsxbb，加入读者圈，一起讨论最新鲜的人工智能科技哦～）
