I started to learn machine learning in three years. When I first to learn machine learning, I do hardly to make it clear that what does machine need to learn. Actually, as a branch of artificial intelligence, machine learning is an algorithm to acquire series of parameters of the function that you want to get. We can regard a system that have an input and an output as a function, e.g. y=f(x). All what we need to do is find a function to map x to y.
With machine learning, a large number of applications have developed rapidly, such as computer vision, speech recognition etc. According to the type of machine learning tasks, it can be divided into supervised learning and unsupervised learning, as well as semi-supervised learning and reinforcement learning. That I have some pictures with true label is supervised learning. If I have some data without label but I want to know which classes they belong to by algorithm, it is unsupervised learning. Combining the two ways is semi-supervised learning. When I train a model with a simulation environment and let model adapt it, what I do is reinforcement learning.
There are two Types of Supervised learning: Classification and Regression. A classification problem is when the output variable is a category, such as “red” or “blue” or “yes” and “no”. A regression problem is when the output variable is a real value, such as “house price” or “weight”. [1]
There are two Types of Unsupervised learning: Clustering and Association. A clustering problem is where you want to discover the inherent groupings in the data, such as grouping customers by purchasing behavior. An association rule learning problem is where you want to discover rules that describe large portions of your data, such as people that buy X also tend to buy Y.
An example to supervised learning
There is a saying in China: “Seeing the morning glow people does not go out, but seeing the sunset it doesn’t matter.” This sentence says that if there is a morning glow, then it is very likely that it will rain, so don’t go out, and if there is a sunset, then you can go out and it can not rain. Why did the ancients associate the sunset glow with the rain?
This is the result of long-term observations by the ancients based on natural phenomena. Although due to scientific and technological conditions, the ancients did not know the causal relationship between them. Through the long-term observations, the ancients found that the morning glow often result in rains, and the sunset glow often result in sunny, then we can draw this conclusion.
If we regard people learning it as a machine learning model, what the model learnt is the relationship between morning glow and rain, as well as between sunset glow and sunny.
An example to unsupervised learning
As the saying goes: “Things are divided into classes, people are divided into groups.” Clustering is also an algorithm of machine learning, which belongs to unsupervised learning. For example, we shop online, and every user has their own preferences. We can record these information and do some recommendation by discover the products they are potentially interested in. After I buy some foods, the machine learning model knew I like to eat snacks so that shopping website maybe recommend more food item to me.
An example to reinforcement learning
Sometimes, when we play Minecraft(A computer game), we need to kill hostile creature or escape from them to protect ourselves and our villagers. We put a machine learning model into games and make a goal to live longer and avoid being killed. If an behavior result in death, the model will be punished otherwise will be rewarded. After that, we could get a model that has a strong ability to survive in Minecraft. It almost always make right decisions to live and avoid death. We can set other goals to activate the model to do a lot of behavior such as build some buildings and planting crop etc.
Summary
In this blog, I have explained some basic theories on machine learning such as what does machine learning do and several concept in machine learning, and give three examples to tell you what the machine learning can do in our real lives. I hope this blog can be helpful to you and have stimulated your interest enough in machine learning.
Reference
[1] Ayush Pant. Introduction to Machine Learning for Beginners[OL]. towards data science. https://towardsdatascience.com/introduction-to-machine-learning-for-beginners-eed6024fdb08