In recent years, due to the development of technology, machine learning and deep learning have a great impact on our lives and society.
Google Translate, Siri, and facial recognition on smartphones are examples of this impact.
Do you understand the characteristics of machine learning and deep learning and use them properly?
This area can also be confusing with a lot of information, so it’s important to understand and use each correctly.
Therefore, in this article, we will introduce how to use machine learning and deep learning properly, their characteristics, and what they can do.
Also, in the second half of the article, we introduce recommended programming languages for utilizing machine learning and deep learning, so please be sure to read to the end!
Table of contents
- To distinguish between machine learning and deep learning
- Understand machine learning and deep learning
- Indicators that distinguish between machine learning and deep learning
- ① Want to get results faster
- (2) Can a large amount of data be prepared?
- ③Is the hardware spec high?
- ➃ Do you need accountability?
- What machine learning can do
- (1) Equipment abnormality detection
- ➁ Traffic control
- ③ Stock forecast
- What Deep Learning Can Do
- (1) Image recognition
- ➁ Voice recognition
- (3) Natural language processing
- ➃ Automated driving
- ➄Medical Diagnosis
- Python is better suited for machine learning and deep learning
- ① Extensive library
- ➁ Easy to learn
- ③ Large community
To distinguish between machine learning and deep learning
As a first step in distinguishing between machine learning and deep learning, I will briefly introduce each definition.
First of all, machine learning is a technology that learns multiple rules and patterns based on given data, and makes classifications and predictions. It is known as one of the classifications of artificial intelligence (AI) and is used in various fields.
On the other hand, deep learning is a type of machine learning that uses a learning model called a neural network that imitates the neural circuits of the brain. It consists of three layers, an input layer, a hidden layer, and an output layer, and this “hidden layer” is many layers deep, so it is also called “deep learning”.
Therefore, it can be said that machine learning and deep learning are not completely separate things, and it can be said that the recognition that deep learning is included in machine learning is correct.
Understand machine learning and deep learning
And from here, in order to understand machine learning and deep learning in more detail, we will introduce the differences in their learning methods and characteristics .
Machine learning and deep learning learning methods
Machine learning methods can be broadly classified into three categories: supervised learning, unsupervised learning , and reinforcement learning .
On the other hand, deep learning has a large number of learning methods, and new ones are appearing one after another, but representative ones include CNN (convolutional neural network) and RNN (recurrent neural network) .
Characteristics of machine learning and deep learning
The decisive difference between machine learning and deep learning is whether humans process data in advance and make it learn .
Machine learning is characterized by calculating, processing, and inputting numerical values (feature values) that represent the strength and weakness of the features to be processed so that the machine can easily learn, rather than having the data at hand suddenly entered by humans . .
Deep learning, on the other hand, can automatically extract the optimal feature values, so there is no need for humans to process the data in advance.
Based on this, many people think that deep learning should be used. However, there are situations where using machine learning is more beneficial, so in the next chapter, we will introduce indicators that can be used for each.
Indicators that distinguish between machine learning and deep learning
The following four indicators distinguish between machine learning and deep learning .
- Want faster results?
- Can you prepare large amounts of data?
- Are the hardware specs high?
- Do you need accountability?
I will explain each.
① Want to get results faster?
Machine learning is mainly used when you want to get results faster. Machine learning requires less computing power and learns faster.
On the other hand, deep learning has the disadvantage of requiring a long time because it processes more data.
( 2 ) Can a large amount of data be prepared?
Deep learning is mainly used for projects that cannot be processed by conventional machine learning and where large amounts of data can be prepared.
This is because deep learning requires a lot of data to perform complex processing.
Machine learning can respond by inputting the data when there is a limited data set structured by humans, but it has the disadvantage that it is not suitable for complex processing.
③Is the hardware spec high?
Deep learning can be used if the hardware specs are high. Because deep learning performs complex processing, it requires higher specifications than machine learning.
The specifications here refer not only to mere memory capacity and processing speed, but also to the processing parts within the computer.
Conventional CPUs, which act as the brains of computers, are not good at parallel processing in machine learning, so it is important to have GPUs, which are good at parallel processing.
However, it also has the disadvantage of being expensive, with some high-performance GPUs costing more than 1 million yen.
➃ Do you need accountability?
Deep learning, in which a machine automatically extracts features, is convenient, but it has the disadvantage of the so-called ” black box problem ,” in which it is not possible to explain ” why a decision was made .”
Especially in fields where human lives are involved, such as self-driving cars and medical diagnosis, which will be introduced later, the inability to fulfill accountability leads to major ethical problems.
From this point of view as well, the decision to use deep learning should be made carefully.
And from the next chapter, we will introduce what machine learning and deep learning can do and examples of their use.
What machine learning can do
Machine learning can do three things :
- Facility anomaly detection
- Traffic control
- Stock forecast
I will explain each.
(1) Equipment abnormality detection
Machine learning, which is good at learning and classifying multiple patterns, is used to detect equipment anomalies.
Since equipment abnormalities are rare, there are many cases in which the machine learns data within the normal range rather than learning abnormal data with a small number of samples.
Then, the machine detects values that deviate from the normal data (outlier detection) to find abnormalities in the equipment.
Recently, Hitachi, Ltd. has announced that AI will perform some of the inspections of hydroelectric power plants in 2023.
➁ Traffic control
Traffic control is the management and control of traffic volume in order to prevent road congestion and danger. Machine learning is also good at making predictions from given data, so it is used for this traffic control.
For example, based on real-time traffic data, it is possible to minimize congestion by predicting the route and time required for each vehicle to reach its destination and optimizing signal control.
In fact, in a test in downtown Pittsburgh, the system reduced travel time by up to 25% and idle time by more than 40%.
In Japan, efforts toward commercialization are progressing, such as the success of a traffic control demonstration experiment conducted by Sumitomo Electric Industries and the New Energy and Industrial Technology Development Organization (NEDO) in 2022.
③ Stock forecast
As mentioned above, machine learning is good at calculating predictions from data, so it has a high affinity with stock prediction.
For example, by having machine learning learn from past price trends and current economic conditions, it predicts the price trends of stock prices, which fluctuate in real time, and buys and sells at the optimal timing.
Recently, Green Monster, which operates an investment app, has installed a prediction function for stock prices in its smartphone app, and its use is expanding.
What Deep Learning Can Do
Deep learning is one of the elemental technologies of machine learning, so here we will introduce the fields in which deep learning technology excels among machine learning technologies.
Deep learning has five areas of strength:
- Image recognition
- Voice recognition
- Natural language processing
- Medical diagnosis
I will explain each.
(1) Image recognition
Image recognition is a technology that identifies features from an image and identifies each part that makes up the image.
In machine learning, when recognizing images, humans must set in advance ” where to focus on data ” (setting of feature values).
However, for complex data such as images, it was difficult for humans to set feature values.
Therefore, by using deep learning, which can automatically discover appropriate settings, it has become possible to recognize images using features of each data that humans cannot notice.
Image recognition technology that utilizes deep learning is also used in face recognition systems. Recently, from March 2022, it will be possible to enter and pay at the Tokyo Dome using face recognition while wearing a mask.
In this way, deep learning image recognition technology is evolving even due to the corona disaster.
➁ Voice recognition
Speech recognition is a technology that recognizes words spoken by AI.
As with image recognition, deep learning is technically more suitable than conventional machine learning because the machine automatically discovers ” where to focus ” in the audio data.
To give a familiar example, this voice recognition technology is used in the smart speaker “Amazon Echo” and Apple’s “Siri”.
(3) Natural language processing
Natural language processing is a technology that allows computers to process languages such as Japanese and English that humans normally use, and to perform syntactic analysis and semantic analysis.
Unlike the image data and audio data mentioned earlier, human language is not originally digitized.
The ambiguity of the language therefore makes computer processing more complex. That is why deep learning, which is suitable for more complex processing, is used.
In our familiar example, it is used for machine translation such as Google Translate and DeepL, and even character conversion prediction.
➃ Automated driving
Autonomous driving means that a computer performs ” recognition “, ” judgment ” and ” operation ” in driving. Among them, deep learning is closely related to ” cognition “.
Specifically, it uses the image recognition technology introduced earlier to identify whether the people around the car are humans or objects, and plays a role in making subsequent “judgments.”
Recently, many companies have released cars equipped with this automatic driving function.
In addition, Nissan has announced that almost all vehicles sold will be equipped with autonomous driving functions by 2030, and autonomous driving technology will continue to spread.
Deep learning techniques are also frequently used in medical diagnosis.
For example, image recognition is used in X-ray and MRI diagnosis.
In 2021, Google launched a trial service in Europe that can diagnose skin diseases by sending photos taken with a smartphone and answering multiple questions as a technology that can be examined outside of hospitals.
The AI market in the medical field is expected to continue expanding around the world, and is expected to reach 7.3 trillion yen in 2027, about 18 times the 2019 level.
Python is better suited for machine learning and deep learning
Python is recommended for those who want to learn AI-related programming such as machine learning and deep learning.
There are three reasons for this.
- Extensive library
- Easy to learn
- Large community
I will explain each.
① Extensive library
Since Python is used all over the world, a wide range of scientific computing libraries have been created.
In addition, the artificial intelligence library is also substantial, so it has a high affinity with machine learning and deep learning.
➁ Easy to learn
Python was developed with the goal of being a simple language, so the amount of code is small.
In addition, the writing rules are fixed, and the grammatical rules (indentation) are always used, so it is easy to read even for beginners.
Furthermore, Python does not require compilation (the process of rewriting the written program into words that are easy for computers to execute). For beginners, it can be said that it is a language that is easy to learn smoothly because it saves time and effort.
③ Large community
Python is currently the most popular language in the world.
Therefore, discussions within the development community are lively, and it is easy to search for and solve problems.
For example, famous Python communities in Japan include ” Python.jp Discord server ” and ” PyLadies Tokyo ” for women, so feel free to join them.
In this article, I have explained the proper use of machine learning and deep learning.
First of all, it is important to understand the advantages and disadvantages of machine learning and deep learning.
On top of that, proper use of machine learning and deep learning is the key to project success.