[PYTHON] An introduction to machine learning
What is machine learning?
Hello! I am a university student majoring in a university liberal arts course in Tokyo.
I've heard the word AI, but who do you think you can actually do?
Actually, I think many people think the word AI vaguely!
This time, I would like to summarize machine learning, which is one of the areas of AI!
I think there are some mistakes and some parts that my understanding has not caught up with, so I would appreciate your guidance in that case!
Let's Start!
Assumed reader
・ Those who are interested in AI but do not know how to get started
・ Those who know the word machine learning but have doubts about what they can actually do
What is machine learning?
Enter from the definition.
Machine learning is "the application and science of algorithms that understand the meaning of data".
In other words, we understand the "meaning" of the various data we come into contact with in our lives (the size of the condominium in Bangkok and the rent and the length of the Euglena cells), and use it to perform various "applications" ( You can use the size of the condominium to predict the rent, and use the cell length of Euglena to determine whether it is male or female).
* I thought about the example appropriately.
Types of machine learning
Machine learning is roughly divided into three types: supervised learning, unsupervised learning, and reinforcement learning.
What is supervised learning?
Supervised learning is a method of learning a model from training data (training data) in which the answer (correct answer) of what you want to predict is known, so that you can predict unknown data and data that will appear in the future!
And it can be divided into two according to the quality of the correct data.
・ Supervised learning with discrete value properties such as whether the correct label (answer) is rainy or sunny, 0 or 1, male or female, etc. is called "Classification".
・ Supervised learning with the correct label having the property of continuous values such as rent and temperature in Bangkok is called “Regression”.
There is no doubt that this method works very powerfully for data for which past correct answers can be prepared in advance (sometimes that is not the case).
What is reinforcement learning?
Reinforcement learning is a method of developing agents that improve performance based on interaction with the environment.
Information about the current state of the environment also includes rewards.
In other words, the environment rewards the actions taken by the agent, and the agent who wants to receive more reward enhances the performance!
The agent who works hard is cute.
Unsupervised learning
In supervised learning, correct answer data was prepared in advance when training the model.
In reinforcement learning, the degree of reward was prepared for the behavior of the agent.
However, the actual society is not so sweet. There are times when you want to come up with an answer by doing something that you don't know the answer to.
Unsupervised learning deals with unlabeled or unstructured data.
With unsupervised learning, even if you don't know the answer or have no reward, you can retrieve meaningful data by examining the structure of the data.
For example, in unsupervised learning called Clustering, by putting together data that are close together,
You can make several groups of close ones.
Get to know the three learnings
I hope that many people have changed their perceptions by learning about the vague thing of AI and the slightly more realistic thing of machine learning.
Machine learning is also a very powerful tool if you master it, but on the other hand, my impression is that if you do not have the knowledge of the user, you will have a child with poor performance.
I will write an article again.
Also, I want to use more photos!
Thank you for reading this far.