Although I knew that the machine learning course in the example was good, I continued to go through it for two years, but it was still good

The other day, I completed the "Machine Learning" course on the online learning site Coursera. This was the best study, but for those who are interested in machine learning and are starting to collect information, the story " Coursera's machine learning course is recommended </ b>"

"Yes, I know </ font>"

Isn't it like </ b>?

(For example, if you search on Qiita, you will find the following very popular articles related to the course)

-The shortest path for working programmers who have avoided mathematics to start studying machine learning --Qiita -Result of studying machine learning for one month from zero --Qiita

I think that's what I've been doing for about two years, while being recommended to the same course by people and articles many times.

However, about two months ago, when I started this course for some reason, I think it was the best as it was well received, and at the same time, I knew that this course was good, but through I was wondering if there are many people who continue to do it </ font>, and while listing the "reasons for not doing" </ b> that were in my brain (albeit noisy), "No, <b. > Still, you should do it </ b> "."

"Stanford class" level seems high

No matter how much you say "Introduction" to the class at Stanford University, the highest peak in the world

"It's an introduction at a high level, isn't it?"

I thought. Even if you read a technical book, you won't understand the feelings of someone who doesn't come to your mind at all.

No, this class can be understood by monkeys.

... I can't say </ font> </ b>.

After all, a lot of mathematical formulas come out and I don't understand what I'm saying. I feel sleepy. But at least you don't need the world's best prerequisite knowledge </ b> to enroll in Stanford.

If you can do high school level math and programming, it is enough as a prerequisite knowledge, and Dr. Andrew is crazy to explain the meaning of mathematical formulas. From the point where the terms are cut off and simplified, they explain the nature of the formula in order and talk about how it is useful in machine learning, so there is a sense of conviction and hunger. ..

And sometimes, " Even if you don't understand the meaning of this formula, Machine learning can be done using a library implemented by a great person </ font>." I will also tell you. Everywhere I feel that I am a teacher who understands the feelings of people who do not understand.

Learning in English about a field you don't understand ...

There are Japanese subtitles </ b> in the whole story. It's not the unpleasant Japanese of automatic translation, but the natural Japanese by volunteers.

At first, I felt like " watching English videos with Japanese subtitles is something I lost </ font>" "Watching in English also helps me study English, two birds with one stone" There was also a strange commitment, but

--The first decision is to learn machine learning ―― (By watching in English) If the psychological hurdle becomes high and you stop taking the course, you will fall down.

I reopened and watched everything in Japanese.

Since then, various online courses have appeared, and I think there is something better.

I don't know when this course has been around, but I think it was already in 2012. Nowadays, some courses are explained in Japanese by Japanese people from the beginning, some cover deep learning, and some are highly evaluated.

One of the great things about this course is that Dr. Andrew's explanations are easy to understand and practical, but of course they are also good.

Weekly programming assignments are required </ font> </ b>

In that respect, I thought it was overwhelmingly better than other online courses.

You need to implement yourself </ b> the neural network or recommendation system you learned in the video and submit it weekly [^ 1].

[^ 1]: It is automatically graded immediately

This is really great

――Even if you don't understand well by watching the video, you can understand it well by writing the program yourself. ――Even if you think you could understand by watching the video, when you write the program yourself, the part where the understanding was ambiguous may be highlighted </ b>. ――You can get the confidence that " I can implement this anymore </ b>"

I felt that it had the effect. I haven't seen this format much in other online courses, and I even think that no matter how good the video is, it doesn't make sense without it [^ 2].

[^ 2]: Except for those who can move their hands and learn without being charged as homework

By the way, recently, the book "Deep Learning from scratch" has been very well received, and I think many people have started / want to learn from scratch.

Deep Learning from scratch-the theory and implementation of deep learning learned in Python
posted with amazlet at 17.02.07
Yasuki Saito
O'Reilly Japan
Sales Ranking: 65
See details on Amazon.co.jp

I recently bought it and read it, and thought that this Coursera machine learning course also has something that is similar to the good concept of "moving your hands and implementing it yourself". As mentioned above, you will implement everything you have learned </ b>.

By the way, I personally think that this book and Coursera's machine learning course (although there are some parts to be covered) cover different areas, so there is no loss in doing both </ b>. The Deep Learning book feels like learning the elemental skills / knowledge to reach deep learning in the shortest distance, while the Coursera Machine Learning course feels more comprehensive about the basics of machine learning. [^ 3]

[^ 3]: Personally, I thought it would be easier to understand if I did Coursera first and later in this book (skipping what I already know).

I want to do it in Python instead of Octave

Many people may avoid it when they ask, "If you're doing machine learning now, it's Python, why Octave ...?"

Of course, I also touched Octave for the first time, and I think that I will not touch it anymore other than this course, but there was almost no overhead of learning Octave's unique usage and notation </ b>. think. There is a video of how to use Octave in the early stages, but I skipped most of it and googled it every time I needed it for testing and assignments.

The implementation I did in the programming task I understand the contents myself </ b> (what kind of function is used and why I am doing such calculations), so I will < font color = "Red"> I think it's easy to port to other languages and frameworks </ font> </ b>.

I'm busy right now, so calm down ...

Some of you may have read this far and thought, "It looks good." But at the same time, I guess there are many people who think, "I'm busy right now, so I'll do it when I'm calm." I have work every day, and engineers have a lot more to learn.

I was the same. I was very busy with my work and private life, and I had other skills that I should prioritize to learn.

"It should be calmer than it is now in about a month, so around that time ..."

I thought, but suddenly I changed my mind when I saw the curriculum of the course.

  • https://www.coursera.org/learn/machine-learning#syllabus

I thought "I want to understand ** neural networks ** properly", but in this course, I learned about neural networks at WEEK 4, 5. Assuming that you started taking classes one month later, you will be able to proceed properly and learn the basics of neural networks one month later.

While this machine learning / deep learning was booming, I was shocked that ** I wouldn't even be at the starting point for at least two more months </ font> **. I decided to start right away.

At first, I didn't focus on completing it, so I suddenly thought about doing only the neural network chapter. However, the curriculum of this course is ** designed to build up knowledge **, and the knowledge learned in the early stages of "linear regression" and "logistic regression" has been relevant to all parts of subsequent chapters. So I'm glad I did it from the beginning.

Tips to continue

I recommend it this way, but I don't have any specific plans to do anything with machine learning, and my work and private life are usually busy </ font>. In fact, I think it's quite difficult to finish the race.

I thought that the first week would be something like "History of Machine Learning" that I could take a look at, so I started casually at first, but suddenly the formula was suddenly changed depending on whether it was the first video or the second video. I remember having a hard time coming out. I'm usually heavy </ b>.

Even so, I think there are three reasons why I continued, so I will write it for reference.

1. There was an intermediate goal

At first, I didn't intend to finish the race, so I thought I'd do it until the 5th week of neural network.

So, I did it until the 5th week, and I didn't want a certificate of completion, so I could have stopped it, but there were some items I wanted to learn, such as "dimension reduction" and "recommendation system" in the curriculum after that. I wondered if I should try it a little more.

When I did it until about the 8th week, I noticed that the last 10th and 11th weeks had no programming tasks </ font>. If you do your best for another two weeks, you'll finish the race! When.

It is a story that I was able to finish the race while looking at the intermediate goal.

2. There was an awareness that "if you get frustrated, you will not be able to study online for the rest of your life."

Due to my nature, once I set a precedent for breaking my will, it's no good </ font> </ b>. That becomes the standard.

"I'm very busy this week, but if I skip even a week, I won't be doing this course anymore."

"If you don't finish this course, you won't finish any online course for the rest of your life </ font>."

It may seem extreme, but since I've been living for nearly 40 years and it's actually like that, "I can't afford to lose the valuable learning option of online courses here </ b>." With that in mind, I was proceeding in the second half.

3. I proceeded while flying moderately

For example (as mentioned above) I skipped the Octave writing video, skipped all the textual commentary and PDFs in the first half, and mercilessly skipped the "optional" problem of programming tasks. There is no doubt that you should do it, but it is better to move forward than to frustrate.

Summary

That's why Coursera's "Machine Learning" is recommended!

  • https://www.coursera.org/learn/machine-learning

Recommended Posts