How to use machine learning for work? 01_ Understand the purpose of machine learning

Hello everyone. I started Qiita from today.

As a machine learning engineer, I am usually involved in AI development and web application development.

In my private life, I became a new dad. I am writing an article while holding my eldest son who is 4 months old (laughs)

As a motivation for writing articles, I have been working on machine learning since 2017 when there was no reference book yet, I feel that the experience of learning machine learning through hard work is definitely useful for someone.

Therefore, I decided to write an article that can support people who are aiming for career advancement. The theme of the memorable first post was "** What was machine learning made for?" **"is.

No matter how good a reference book you read, you can't learn anything unless you know the purpose of machine learning. First, let's understand the big picture that is the basis of everything.

We also send various information on SNS, so if you feel good reading the article I would be grateful if you could follow Twitter account "Saku731".

What is machine learning?

I think everyone has had such a hard time.

** "I don't know what I learned from machine learning" ** ** "Logistic regression, decision tree, support vector machine, there are too many methods" **

One of the pain points in learning machine learning is ** many techniques to remember **. It is necessary to understand 5 to 10 methods just by reading one reference book.

What's more, ** "when" and "how" to use those methods properly **? I have never come across a well-explained reference book.

Then, on the contrary, why not understand from ** why there are various methods ** and their ** commonalities **? That way you can get a faster overview of machine learning.

A purpose common to all machine learning. It means making a ** good model **.

As you can see in the figure, this is all about machine learning. "** Give a large amount of data to a computer and find out the rules for making predictions **"

However, there are various types of data in the world.

--CSV (Excel): Sales data, customer data --Image: Photos posted on Instagram, images from surveillance cameras --Voice: Meeting voice memo, call center call recording

Machine learning engineers are required to create good models (highly accurate predictions) of any data. Various methods are prepared as a tool box for that purpose.

--Sales data ⇒ Regression (multiple regression, regression tree, etc.) --Instagram photo ⇒ CNN-based method (VGG16, LesNet, etc.) --Voice memo ⇒ Time series analysis (RNN, LSTM, etc.)

In the case of cooking, the common purpose is "** to make delicious food **". For that purpose, it is the same as using the appropriate ** recipe (= method) ** according to the ** material (= data) **.

Organizing terms related to machine learning

Now that you can see the purpose of machine learning, let's broaden our horizons a little more.

As the first struggling part of machine learning ** AI / Machine Learning / Deep Learning ** I get the impression that many people do not understand these differences.

If you read the news and articles of 2018 when AI (machine learning) became popular, about these three I remember that there were many sentences that made me feel, "Isn't the person who wrote it understand too much?"

If you don't understand the difference here, even if you study hard This is an important issue because it is not systematically organized as usable knowledge.

First of all, the three terms ** AI / machine learning / deep learning ** have a comprehensive relationship as shown in the figure.

(Source: https://www.shikaku-square.com/media/ai-license/001-how-to-study-deep-learning-for-general/)

As you can see, it is a comprehensive relationship of "AI ＞ Machine learning ＞ Deep learning". It is as follows when organized in words.

• AI ( Artificial Intelligence )： --The concept that was born in the first AI boom of the 1950s --The purpose is to artificially reproduce human intelligence with a computer. --Machine learning: --A concrete method that was conceived to realize AI --The purpose is to find out the law from a large amount of data and predict the future. --Deep learning ――Among machine learning, the field that uses the technology ** neural network ** -The purpose is to automate the most important (and laborious) AI development process called ** feature extraction **

By valuing such basic knowledge, you can deepen your understanding of articles and reference books at once. Also, when it comes to machine learning projects, it becomes a knowledge that even non-programmers should know.

at the end

Next time, after organizing the ** AI development project overview ** I will write ** "where" and "how" ** where machine learning is used.

• (Added on December 24, 2019) "2nd: Overview of AI Development Project" has been posted.

If you have studied machine learning, you may have heard of it. ** Each topic such as "pre-processing" "learning" "tuning" "verification" ** It will be organized in association with the progress of the AI development project.

In the next article, it will be based on the programming language ** Python ** We will organize specific study methods, so please look forward to it.

• (Addition) "3rd: Full disclosure of Python coding procedure" has been posted.

Thank you for reading for me until the end.

【P.S.】 We also send various information on SNS, so if you feel good reading the article I would be grateful if you could follow Twitter account "Saku731".

~~ Also, at the end of the sentence, we are doing "** Team Development Experience Project **" for a limited time. ~~ ~~ If you are interested, please check [Application Sheet] for details. ~~ (Addition) The deadline has been closed because it is full. The next time is scheduled for March 2019, so if you would like to be informed, please fill in [Reservation Form].