Mastering Machine Learning with a MOOC Portfolio

If you don't study hard today, you'll try to find a job tomorrow


In January 2017, humanity's arrogance, built up over tens of thousands of years in the intellectual realm, was mercilessly suppressed in several ways.

Let's not mention the exciting self-driving products at CES 2017 (I guess you can throw away your driver's license in seven or eight years) or go into detail about the human elite being crushed in the image recognition segment of the show "The Best Brain" (which used to be the quintessential area where humans could laugh at AI), but let's talk about the tsunami in the Go world.

Dozens of top human Go players, using the millennia-old wisdom of human Go play, struggled hard and were still beaten by Master with a fast game - 60:0.

Others were bewildered after the loss, but only Gulli's words after the battle were the most intriguing - "The curtain is about to open for humans and AI to explore the world of Go together".

That sounded extremely familiar to me in rhyme. It suddenly occurred to me that a similar episode was described in Wang Shuo's The Tenacious Master.

Ma Qing strode forward with gusto, shaking his fists at the pedestrians and barking, "Who the hell dares to mess with me? Who the fuck would dare mess with me? "A big, burly man in overalls approached him and whispered, 'I dare you to mess with me. "Ma Qing froze for a moment, sized up the pylon of a lad, and looked around and said, 'So who the fuck dares mess with us two? "

Jokes aside, it's good to see the human elite finally learning humility and expressing a sincere desire to cooperate with AI. The rise of artificial intelligence has made many people think they recognize the reality that

Not when you have work today. do artificial intelligence (AI), Tomorrow artificial intelligence will take your job。

Do you agree??

Actually, that's it mistaken It is. not Reality.

Man and machine don't compete.。 People earn money to eat., The machine has power.。

The reality is.

You don't learn AI today when you have a job, tomorrow anyone Using artificial intelligence to replace your job.

People and people will never be able to compete. In this arms race, artificial intelligence has become a nuclear weapon that individuals can master. People understand, you don't understand, the results will be wonderful.


Of course I'm talking about other person results。

In the field of artificial intelligence, The hottest technology at the moment is the machine do(machine learning), In particular, the rise of big data in recent years has been accompanied by the creation of depth do(deep learning)。 whetherAlphaGo Or Baidu Brain, It's all actually used( depth) robots do, That's what makes it so great.。

Would you like to learn about machine learning?


may understand and mastering the machine do, There are many paths。 You can take classes.、 enroll in a class, Even read the book yourself。 But what I recommend more, beMOOC。

For most people,MOOC Easier access to resources。 As long as you have a computer with internet access, You can start learning。

Because these years the machine do very hot, Therefore the relevantMOOC enough。 Some good people have started to sort out the summaries, Dividing dozens of courses into primary、 mid-level、 high level, Each section has more than ten courses。 And then list them., For readers to follow the steps to do。

My assessment is that they have done a very serious and responsible job with a terrible what happened。


Because you're listing dozens of courses in one fell swoop, you're not lowering the bar of learning in any way, but rather quickly reducing the number of "introductory to give up "The length of time.

sure!, After reading your list., readers on give up finish。 His conclusion was“ robots do This doesn't seem to have anything to do with my life.”。

This article attempts to do something different - a select few courses that assemble a path that is easy to get started and quick to apply.

To achieve this goal, it is necessary to select the "best" course from the many MOOCs available. The so-called "best" courses need to meet the following conditions.

  1. (a) The teachers themselves are cattlemen.
  2. Teaching intentions;
  3. The curriculum is designed to meet do rule (e.g. of science)。

With this standard, Many courses were immediately ruled out。 We have selected several courses, Lead the beginner through each of the entrance door、 advanced、 Applications and Lessons et al. process。 put together, It's a complete machine. do understand and mastery of the path。

Because it's hard to adjust to the crowd, So it's impossible for us to find out that which anyone (not) at all recognized“ preferably” programmes。 If you don't think the courses I've listed or the way they're combined are good enough, Don't just start cursing.。 Because of this not roundPK。 You are welcome to recommend a better course to us all, thanks!

entrance door

entrance door Course recommendation Wu Enda(Andrew Ng) lecture onCoursera on the platform《 robots do》。

Professor Enda Wu, the man who brought artificial intelligence to smash the show in The Greatest Brain.

He is a professor at Stanford University and one of the co-founders of Coursera. He is now the Chief Scientist of Baidu and has overall responsibility for Baidu Research Institute.

If you post a question on the MOOC forum, the webmasters give you recommendations for first door robots do It's probably the same course.。

The reason is that the threshold is low and the ceiling is high.

So-called low threshold, It means that you are not required to be very proficient in a certain language programming technique, Nor do you need to have the mathematical knowledge to do a number of volumes《 Jimmy Dovich Mathematical Analysis Workbook》 atmospheric level。

The so-called ceiling is high, means that the course provides a comprehensive introduction to machine do Core knowledge in the, and extends to a number of different applications。 In fact, I really think this approach to curriculum design is very worthwhile for our universities to do。 While ensuring coverage, Only focus on a few key knowledge points and practice them repeatedly, Makes it easy for students to learn, But he's really got the essence.。

I think that's the true level of mastery.

A previous article was written specifically for this course《 robots do, You can do it.》。 I won't go into details here。 Interested students are welcome to go out and turn left to check it out。

After taking this course, you can proudly proclaim yourself understand Machine learning now.


After receiving a certificate from Professor Enda Wu's course, Never strike while the iron is hot., Get to grips with today's hottest depth do。

so-called depth do, It is the use of depth Neural networks for machine do。 The figure below shows the use of depth Social networks to identify a picture is actually a car、 human, Or an animal.。

The recommended courses here areGeoffrey Hinton lecture on《 Neural Networks and Machines do》。

Why do you recommend this course?

firstly, You will find that this course uses the same software tools as the previous course, All are easy to install、 handyOctave。 You don't have to go understand Various configurations of the programming environment, Just get it and use it.。 The framework code is provided, You only need to make changes and additions in the key areas。

Secondly, the course defaults to you having taken Professor Ng's course as a precursor. Thus there is little overlap between the two to avoid duplication of effort. The direct high level of the house continues to advance in depth.

thirdly,Hinton The professor is depth do A recognized authority in the field。

This great man's greatest skill is that he has his own thoughts and opinions about his career and does not follow the crowd. When the AI field hit a cold winter, the grants were gone and the labs couldn't sustain themselves. Many talented people jumped to other fields, but his old man was the only one who always felt that there should be opportunities in constant improvement and development, and stayed.

The rewards of this persistence have proven to be significant. He was decades ahead of everyone else in the deep end, and scholars today can barely get around citing his literature in whatever new papers they write in the field of deep learning. :-P

So, it's really a blessing to hear such a titanic figure teach you. A very esoteric algorithm to others, which reads either like a heavenly book or is admired, but in his mouth it is nothing more than "Oh, I tried this one year and it didn't work; then suddenly I got enlightened and changed it a little here and there ......"

Some of the people in discussions The district left a comment saying his jokes were so funny。 It's true.。 But don't expect too much from this thing when you learn it—— As a foreigner, You live in a different environment and have a different cultural heritage, Don't expect to understand all the jokes.。

After taking this course, it will be very difficult for people to try to fool you with the concept of deep learning.

Did you get that?

After taking this course, If you take depth do The concept of fooling around with other person……


put into practice

Through the first two courses, You understand what a machine is. do harmony depth do。 But you may not want to stop at just a conceptual understanding, But I'd love to make something to practice.。 Practical applications on the one hand can test your do effects, On the other hand it can help you gain experience, It also brings a sense of accomplishment。

Sorry, there's a branch in the path here in the application practice session.

The first option isUdacity The course above"Intro to Machine Learning"(ud120)。 The lecturer isSebastian Thrun, For the first few years he had beenGoogle Head of self-driving technology。

The function library used in this course isScikit-learn, built onPython above。 The course is designed to be engaging, Using some data sets, To familiarize you with how to use other person Well-constructed features( Function calls), Add your own understanding( Parameter adjustment) to complete some small projects。

I was generally happy with the course, The trolling is on the free do user-oriented, For lack of necessary support, The final big project is steeply more difficult。 Hard for the average person to adapt, It's easy to get halfway through give up。 Of course, If you are a paid user, There will be someone to lead you step by step, and provide the necessary help, I feel much better。

The second is choiceCoursera platform at the University of Washington machines do Course Special Series(specialization), The course name is“Build Intelligent Applications”。

To start with the spiel, I think this course clearly reflects the interference of capital markets in academic activities. The course series consists of six courses. But so far only the first 4 doors have been released. Gate 5 has been bouncing around constantly. It was delayed from September 2016 until the end of January 2017.

Why? Because the keynote speaker's two companies, Turi, have been Apple acquisition Up. So they had to hold off on the production of the course at hand and deal with the various aspects of the acquisition.

However, judging from the four courses that have been released, the standard is very high and cleverly designed.

The course uses case teaching, Is that each module is processed according to a given data set for a specific problem。 first door The lesson is an overview, Let users experience the important content of each part, And run the try, Increase interest and confidence。

Then starting with the second course, each course introduces a separate machine learning type. Each week's lessons are progressively ramped up from light to deep. Use Jupyter Notebook to save the results of the exercise, which can be used directly in practice in the future.

The couple are also very generous in opening up their company's products to students for free。 Their products are also free for academia( Certificates need to be updated manually and free of charge once a year), Very generous。 But was acquisition later, This policy is not easy to say。

Some chapter topics are somewhat difficult。 As do he who, You have to stay away from your comfort zone, It's not stopping do Preparations for a tour of the district。

After listening to one of these two courses above, you should be equipped to work with some small data sets on your own. It should be of great benefit for study, work and research.

You're dissatisfied and want to handle massive data? This matter requires two conditions to support it.

  1. You need access to large-scale data.
  2. You have to have the appropriate hardware.

To tell you the truth, the vast majority of people are now not have The appropriate conditions. So we'll talk more about this question when we're free.


You may not have much confidence in your knowledge base—— I not Mathematics or computer-related majors, You can also learn machines do Is it??

From the course just now, You shouldn't be hard to see。 Now the machine do The threshold is already low。 But in order to achieve the goal, You still need to have some very basic knowledge, Mainly including:

  1. Mathematics (elementary calculus, probability theory and statistics, linear algebra).
  2. Programming languages (e.g. Python)

If you lack the appropriate knowledge, that's okay, just make it up. Fill in where you're missing and add to it repeatedly so that your learning energy and efficiency are guaranteed.

Python linguistic Lessons way I suggest the University of Michigan'sCoursera Course Series《 Everyone can learn to program》。

This is by far the best I've seen for data science beginnersPython language entrance door method。 I took this course, You not only master a language, You can also master the basics of data mining。 Include database operations、API and reptiles, etc。 It can be docked directly to the previous practice class, Tight silk seams。

As for a supplement to your math knowledge, my recommendation is the Khan Academy, which Bill Gates has been raving about.

Khan College's first skill is to help students tutor math。 So inMOOC field, The platform's math curriculum is the most accumulated, It is also the least painful for students。

See so many math course categories at Khan Academy? Don't hesitate to learn. You won't relive the nightmare of your first year of college because the lecture style is really lively and interesting.


What else have you been on to explain machines? do goodMOOC? Where is it unique?? Can you replace some of the courses listed in this article?, Form a better combination? Welcome to share the message, Let's go together discussions。

1、NYK to push its own digital currency to pay salaries
2、IPC Italy steals the show in the driverless era
3、python implements wechat messages sent on a regular basis and sends automatic daily messages to remind matters
4、Nissan to test driverless taxis on Yokohama public streets
5、The potential is unlimited with many highlights

    已推荐到看一看 和朋友分享想法
    最多200字,当前共 发送