What is machine learning?

What is machine learning ?

Machine Learning, sometimes called Automated Learning, is an old terminology dating back to 1960. It is a contraction of the term machine, which refers to computer interfaces, and the word learning, which defines the ability to discover and recognise structuring patterns in a set of data. The appearance of the term Machine Learning and its abbreviation ML tends to dissociate the new models from the automation processes more common in the 20th century.

The difference between ML and automation is based on the need in automation for an exhaustive enumeration of decision-making rules. ML does not need this, as all decisions are extracted from data sets and sometimes from a few directives presented upstream. This allows the ML model to respond to new problems where previous models are either limited or unable to respond. This is particularly the case in highly complex situations or when the problem requires the algorithms to be adaptable. Unlike a standard programme, an ML process can also be adapted to respond to variations in context. This is not possible for an automated system without the contribution of an external expert.

In the Machine Learning sector, the main objective is to design systems that enable computers to understand structures and assimilate knowledge from data. To achieve this, Machine Learning encompasses a vast array of mathematical tools. The Machine Learning field encompasses several fields of science, drawing on knowledge of statistics, information theory, game theory, optimisation, etc. Finally, as a field of computer science, a large part is devoted to the use of algorithms and the concept of computer science.

What is Machine Learning used for?

Machine learning models are varied, but they all revolve around the use of data to respond to a specific task. Generally speaking, the aim is to carry out an in-depth analysis that brings to light new knowledge based on a set of data and a previously targeted objective. To do this, ML models establish relationships, define patterns and construct representative samples of measurable real events.

From an industrial point of view, ML follows a trend that consists of automating processes that are carried out by humans and have become costly, from which we wish to remove individuals. In the specific field of ML, it is information analysis processes that are targeted. Among other things, this virtual analysis enables a better understanding of the data, and can help decision-making by making more or less reliable predictions using the context represented by the data.

Let’s not forget that we are entering the era of Big Data for mankind and industry. This period is defined by the emergence of a data-rich ecosystem with huge volumes of data. To be able to analyse the data despite the increase in the digital ecosystem, we need tools to automate analysis methods, such as those offered by ML. The fact that ML makes it possible to automate the analysis of these datasets is, in my view, a testament to its strength in meeting the challenges of this century.

What are the outstanding performances of Machine Learning?

The emergence and use of ML in daily life is disrupting several sectors and the associated uses are becoming more and more essential for the development of several sectors. In everyday life, Machine Learning algorithms are often highlighted by their ability to reproduce human behaviors such as language, visual recognition, creation, translation, conversation and analytical decision making. Behavior that is strongly rooted as inherent in the human and animal condition. Each advance in this sector undoubtedly raises philosophical and human questions about our condition and can arouse fear.

ML is often defined as a branch of AI (Artificial Intelligence). But in the case of ML, the exercise is often extended to other objectives than simply reproducing animal intelligence artificially. It is also a question of perfecting certain reasoning and supporting others in decision-making. The external human actor is complementary and essential.

It is important to remember that ML is one tool among many others and that it is not limited to the imitation of human processes. The very strength of ML is the ability of its models to extract new information from large volumes of data. Moreover, the finer a task or a question, the more the use of real and rich data improves performance. However, the analysis is made difficult by the time required for the intermediate tasks of manipulation and understanding of the data, especially when the amount of data is gigantic. This is where ML is a great tool, since it allows you to automate most of the data analysis subtasks at a lower cost by a virtual actor.

By its machine condition, the ML model can perform longer, sometimes in a more optimized and finer way than the human. The strength of ML models is also to have the necessary resources to minimize errors of reasoning by parallelizing the analyzes in several virtual experts, which we cannot do so simply with human experts.

Nevertheless, ML models should not be designed directly as competitors of humans but as practical decision support tools. The machine cannot make a judgment outside the analysis of the facts, the human remains the only one able to choose according to the information given. For example, the email spam filter and search engines are two main use cases of ML already present in everyday life as an assistant giving free rein to human choices.

This gives an idea of the tasks that humans can do (sorting their emails) and which we want to delegate to machines and the tasks that humans cannot do because they are limited by their capacity (propose the most relevant content from the vast internet ).

Machine Learning, a large family of algorithms.

As with any area of computer science, there are many ways to achieve the same things in ML. It even happens fairly regularly that two approaches are the same to respond to a problem.

The fact that there are several ways to do Machine Learning, both to respond differently and subtly to problems that vary slightly in the initial conditions, but also to optimize these results, makes the discipline rich and difficult to grasp in its entirety. together.

There are several sub-branches of ML following the learning approaches put forward. But that should not discourage you, each family in ML has its own universe and it is not necessary to master the whole thing. In a blog post, we go over the different approaches and how they differ to meet specific tasks. Finally, a major element in ML concerns the famous theorem with the evocative name “There is no free lunch”, which reminds all ML enthusiasts that there is no miracle solution in ML. Each approach responds to specific issues in a more or less relevant way.

Photo Maxime Macé

Maxime Macé

Simple passionné de thématiques diverses et variées. J’apprécie enrichir mes connaissances dans les disciplines techniques comme l’informatique, les sciences et l’ingénierie, mais aussi dans les domaines merveilleux de la philosophie, l’art et la littérature.

Prenons contact

Vous souhaitez me contacter suite à la lecture de cet article ? Merci de renseigner votre adresse électronique et je vous recontacterai dans les plus brefs délais.

CV Ingénieur développeur informatique

CV Étudiant PhD Artificial Intelligence

Portfolio Artist designer