In recent years, the deep learning method in machine learning, which is rapidly being implemented, has attracted attention. In deep learning, processing that imitates nerve cells in the human brain is performed, and by changing the weight, it is possible to freely perform processing with different properties. This time, I would like to introduce an article about deep learning.
"Machine learning" and "deep learning" that we often hear when talking about AI (artificial intelligence) technology. How are the two different? We will explain in an easy-to-understand manner the mechanism and differences in the fields of application.
Machine learning is a term that broadly refers to learning from machines (machines such as computers and AI). We will instruct the machine to iterate over a large amount of data and train it to find, judge, and analyze patterns.
There are three types of machine learning: "supervised learning," "unsupervised learning," and "reinforcement learning."
"Supervised learning" is a learning method that gives a lot of data to the machine while showing the correct answer at the beginning so that the machine can show the correct answer through training. For example, if you want to distinguish whether an image is a "cat" or a "mouse", you can learn the characteristics of each by giving a large number of images labeled "cat" and "mouse". The machine will then learn how to distinguish between a "cat" and a "mouse" and will eventually be able to give the correct answer even if you give it an unlabeled image.
On the other hand, "unsupervised learning" does not give the machine the correct answer. Give a large amount of unlabeled data to learn the characteristics, overall trends and rules of each data. As a result, the machine will be able to group the data by features, etc. and determine which group the newly given data belongs to.
"Reinforcement learning" is a method of learning the best method by adding a method of scoring the results produced by the machine to the supervised learning method. By rewarding or punishing the machine, you will get better output.
Deep learning is an advanced method of machine learning and can be said to be a type of machine learning.
Deep learning is based on a technology called neural networks that imitates the mechanism of human nerve cells (neurons). This makes it possible for deep learning to derive the correct answer through a process that is very similar to the human cognitive process.
For example, when recognizing an image of a "cat," the machine finds the features of the object, such as the shape and contour of the ears in the image, and matches it with past data to give an answer. However, in the case of human beings, the conclusion is reached through a number of hierarchical processes, such as looking at the whole and then checking the details and vice versa. By doing so, an image of a "bear" that has similar characteristics to a "cat" can be recognized as an animal that is not a "cat" (sometimes it cannot be done).
In deep learning, machines are taught and executed such human-like recognition processes and ways of thinking. And as a result of giving a large amount of data to a high-performance machine for learning, the machine is now able to demonstrate pattern recognition ability that surpasses that of humans.
The difference between machine learning and deep learning
The big difference between machine learning and deep learning is that deep learning has multiple layers of "eyes" when analyzing data. Deep learning also learns the "eye-catching place" itself and improves its performance.
For example, in order for machine learning to recognize colors, it is common to instruct humans to focus on "color", but deep learning does not need to do so. In the case of deep learning, the machine itself automatically learns the characteristics of the data, finds a way to distinguish colors, and obtains the correct answer.
In addition, machine learning and deep learning have different areas of practical use. Machine learning tends to be used to make decisions as directed by a person. Examples of its use are face detection with a camera, spam detection by email, fraud detection using a credit card, number recognition written on paper, and conversation understanding.
On the other hand, deep learning tends to be used to read and judge what humans do not instruct and what humans cannot grasp. Examples of applications include automatic recognition of signs and traffic lights in autonomous driving, pedestrian detection, cancer cell detection in medical research, product appearance inspection, deterioration diagnosis of infrastructure facilities, equipment maintenance and inspection, and automatic translation system of voice data.
Deep learning used for big data analysis
Deep learning is also used for big data analysis, which is difficult for humans. With the advent of deep learning, the accuracy of image recognition and voice recognition has improved dramatically, and the analysis and organization of vast amounts of information contained in big data, image data and voice data is now at the level of being left to AI.
It is said that if deep learning (or machine learning) is used for big data analysis, "predictive maintenance" at factories will be possible. Predictive maintenance is to attach sensors and cameras to devices and equipment, convert them into big data by IoT, analyze them with AI, and prevent breakdowns and malfunctions. AI is responsible for finding signs of failure or failure in big data and alerting when needed.
AI is now evolving at a rapid pace. Understand the difference between machine learning and deep learning and think about how you can use it in your company.