Everyone knows the term "machine learning". But do we know exactly what it is? The technologies this field uses are often unknown, as is the extent of its possible uses.

In this article, we will dive into the world of machine learning and see everything that revolves around it: machine learning professions, artificial intelligence, big data, etc.

Machine learning

The definition

Based on Sleep Spell 5e definition , machine learning is:

[center]A field of study of artificial intelligence which is based on mathematical and statistical approaches to give computers the capacity to "learn" from data, that is to say to improve their performance in solving problems. tasks without being explicitly programmed for each.[/center]

Clearly, and as its name suggests, machine learning is about giving computers the ability to learn , and thus perform actions automatically and intelligently. As we will see later in the article, this learning can be done in several different ways.

In all cases, machine learning has two phases, which can be more or less nested.

First, the learning phase , or observation. It is during this stage that the computer, the program, will learn. We're going to send him huge amounts of data (we'll talk about it again) and ask him to analyze them, to teach him how to solve the tasks that he will have to deal with on his own afterwards. Typically, if we want a machine learning program to recognize cats from dogs, we will send it thousands of photos of each animal and breed of these animals, explaining which is a cat and which is a dog.

The second phase is the start of production . It is from there that the machine is going to have to do what it was trained to do, or what it is self-training for. To take the previous example, we will now send him photos of a cat or a dog without telling him what it is, and the program will know how to make the difference. And, if he is facing an unknown breed, he may be able to find similarities with others he knows, and thus classify him as a dog, or a cat.

Historical

Contrary to what one might think, the field of artificial intelligence and machine learning is not new.

The first to use this expression was an American computer scientist, Arthur Samuel , in 1959, over sixty years ago. This neologism came to him 7 years after he had created a program that played Checkers, a program capable of improving itself over the games.

Ultimately, his algorithm( [pii_email_aef67573025b785e8ee2] ) managed to beat the 4th best player in the United States.

Machine learning and big data

While it depends on the type of learning machine learning is based on, machine learning often needs huge amounts of data to learn on its own . To take the example already mentioned, if we want a program to recognize a cat from a dog from a photo, it must have already seen tens of thousands of each.

The more data a machine has in its learning base, the more accurately it will be able to perform its automated tasks with a low margin of error.

The different applications

Obviously, we don't just use machine learning to be able to tell the difference between cats and dogs in a photo. The applications of this type of technology are vast and much more useful in our daily lives.

Typically, machine learning is present in the field of self-driving cars . It is also found in social networks (generation of a personalities news feed). Or in the health sector (assistance in recognizing anomalies on medical radios, for example).

Concrete examples of the use of machine learning will be given at the end of the article.

The 3 main types of learning

As we said, there are different types of learning that allow a computer to do machine learning. In this section, we will detail the three main ones.

Supervised learning

In supervised learning, the program is provided with explicit, or "tagged" data. For example, if we want a machine to help us differentiate a medical X-ray of a healthy organ from that of a diseased organ, we will provide the system with thousands of images, with, for each, a label saying "organ healthy ”or“ diseased organ ”.

To train it, we provide everything ourselves to the program, hence “ supervised learning ”.

That way, after reviewing those thousands of images, the program should be able to tell the difference between a good radio and a bad radio.

Unsupervised learning

There is an approach completely opposite to the first type of learning: unsupervised learning.

There it is quite the opposite. We provide the machine with a whole lot of data, and it's up to it to find patterns , to create kinds of groups. For example, if we give a machine learning program a database of people suffering from a certain disease, it will try to find common points between these people, find a pattern that could possibly cause this disease.

And, behind, it is the operator who is in charge of analyzing these groups to draw any conclusions.

Reinforcement learning

The last type of learning we are going to talk about here is reinforcement learning. Here, no need to inject thousands of data into a machine. This time the machine will learn on its own by trying.

This is typically the kind of learning that an artificial intelligence uses to learn to play a game. For example, this is how the AlphaGo program learned to play the game of Go. Part after part, the program analyzes the combinations which have the most chances of leading to a victory, all in relation to the opponent's strokes. AlphaGo thus defeated the world champion of the game of Go.

Machine learning in the technique

Now that we know more about machine learning as such, let's dissect it from a technical perspective like Azurewave Devices .

The profession of data scientist

Usually, the person who manages and supervises the machine learning of a program is a data scientist . The data scientist is, roughly, an expert in the management and exploitation of data .

In particular, he knows how to manipulate big data, and use it correctly in order to train a program in the sense of machine learning.

He is able to use the tools he masters (programming languages, frameworks, etc.) to teach a computer to perform tasks automatically and by learning on his own.

Machine learning techs

The data scientist, to do machine learning, can rely on various technologies.

In terms of programming languages, there are mainly two here: R and Python . Since R is more specialized in the study of statistics (exploratory analyzes, correlations, etc.), we will focus here on Python.

This is also one of the explanations for the increasingly widespread use of Python. It is indeed the language of choice for work on artificial intelligence and machine learning.

As a result, there are today many frameworks running under Python whose goal is the exploitation of big data and machine learning.

If you are a developer and want to specialize in these areas, then Python is the place to be.

Concrete applications of machine learning

As we saw at the beginning of the article, machine learning is already present in our daily life, without even noticing it. Let's see some concrete cases of its use.

Personalized flows

Whether it's your Facebook news feed , your Instagram or Netflix suggestions , there's machine learning behind these feeds.

Concretely, we want you to stay as long as possible on the platform (Facebook, Instagram) or that you find what you are looking for (Netflix).
How to do ? Well, an algorithm, trained to generate suggestions that you may like (depending on your previous searches, the time you spent on a particular page, your likes, etc.), will be called and thus return personalized content.

Field of health

As already mentioned in this article, machine learning is perfectly suited to the world of health . It can be of great help in diagnosing abnormalities on analyzes (supervised learning), or finding possible causes for a disease (unsupervised learning).

If this particular area interests you, head over to this article .

Fraud detection

Another example that can be given here is the detection of bank fraud. For example, a machine, by dint of processing data, can deduce from you that you live in such and such a city, with such an inflow of money, and such expenses each week in such and such a place.
It will sort of create a model, a pattern , of your situation.

If this same program sees a transaction for an amount much greater than the usual volumes and in another country, it will identify it as a potential fraud.

Limitations of machine learning

While machine learning can do a lot of things, it has limits on several levels.

Already, and although the term is sixty years old, machine learning( Firebolt 5e ) is still in its infancy. The day we succeed in interconnecting the different learning systems and pushing the limits of this learning, this field will become even more extensive and will disrupt our daily life even more.

Another limit that we can see in machine learning is its ethical or moral side . If a machine is able to create a financial model from your bank account, maybe it can take it a step further. For example, creating different patterns from your personal data, which could be used by third parties for questionable ethical reasons.

Normally, after this article, the term machine learning should be clearer to you. We now know what it corresponds to , what the different types of learning are , what technologies machine learning is based on and what its concrete uses are .

Nice one op, what a simple way of making it look simple. The future is Al there is no way my child will not study data analysis.

viyon02:
Nice one op, what a simple way of making it look simple. The future is Al there is no way my child will not study data analysis.

idiot.. You never even born pikin to know weda him go dey interested in studying data analysis or maybe he has another thing God has destined him for, but your own be say your pikin must study data analysis weda him like am or not. Honestly Nigerian parents are the most useless things on the surface of the earth, steady using their myopic views and their own hands to redirect the lives of their children, making children not enjoy their lives. Being born to Nigeria parents should be classified as a disability.

sqlPAIN:
idiot.. You never even born pikin to know weda him go dey interested in studying data analysis or maybe he has another thing God has destined him for, but your own be say your pikin must study data analysis weda him like am or not. Honestly Nigerian parents are the most useless things on the surface of the earth, steady using their myopic views and their own hands to redirect the lives of their children, making children not enjoy their lives. Being born to Nigeria parents should be classified as a disability.

Your parents must be idiots to channelled your career against your interest, have I told you he most study it as a future career? Basic knowledge of AI is necessary for whatever career he/she decides to study so that he may stand out. However, introducing my child to it never a crime, Infact I will introduce my child to many things, I will leave him to choose whatever interest him, but it will be crime on my part not to introduce him to good things of life. Do not quite me because I won't reply you.peace