"Machine learning" What do you mean ?

Machine learning, is in fact a form of Artificial Intelligence that appeared in the 1950s thanks to the American computer pioneer Arthur Samuel. He invented the term “Machine Learning” by designing an algorithm capable of learning to play checkers but above all capable of self-learning.

Machine Learning can be defined as an Artificial Intelligence technology that allows machines to learn without having been programmed specifically for this purpose. This refers directly to a process of analysis and implementation that allows a computer to realize complex problems as a result of a multitude of systematic processes. There are, however, two main ways of “educating” an algorithm with the Machine Learning technique.

First of all, we have supervised learning: this is the most widely used learning method. It consists in teaching a task to an Artificial Intelligence by reproducing this task many times. By dint of encountering this task several hundred or even thousands of times, the machine will adapt and understand by itself the response expected by the program executor without the need to accompany him.

Then comes clustering or unsupervised learning: in this case of learning, the program does not rely on any external elements. It learns by itself by crossing information and classifying data by similarities. This mode of learning is more complex than the previous one. Its main advantage lies in the fact that it is almost non-participatory and therefore more reliable with less biased results.

Ce sont les deux modes d’apprentissage du Machine Learning les plus courants, mais ils en existe bien sûr de nombreux autres systèmes.

Machine learning: from theory to practice

Nowadays, the learning machine is present in more and more sectors. With the development of new technologies and the growing need for Artificial Intelligence, it appears to be indispensable for our society. Here are a few concrete cases:

Medical field: this learning process has a very important place in the field of medicine. The algorithms implemented are designed to assist doctors in dealing with data flows and to facilitate their interpretation. In particular, researchers at MIT have developed a system for analysing medical images and facilitating diagnosis.


The use of robots in surgery in particular is more and more common and Machine Learning makes it possible to improve their use as operations are performed. Machine Learning is even invited into the research and development process of laboratories generally to help find the tester profiles corresponding to the desired analyses and thus facilitate discoveries.

Automobile: We owe the autonomy of autonomous cars to the automatic learning process. The latter allows us to identify obstacles and do predictive maintenance. All this coupled with the power of computer calculations allows us to analyse the information in real time!

Financial Security: With predictive analysis, financial fraud will be more easily detected and security is enhanced. Thanks to the data collected by the Machine Learning, highly personalized financial instruments adapted to each economic agent will be offered in the near future. The Machine Learning already touches many areas and its development is now reaching many markets.  

Le Machine Learning jusqu’au bout du nez

This automatic learning process is invited in our Serenity Eyewear by Ellcie Healthy connected glasses. In order to avoid as much as possible the over detection (or under detection) of falling asleep, our glasses use Artificial Intelligence and more precisely Deep Learning. This system allows the self-learning of the glasses by means of a network of neurons.

Artificial neural networks function like the neurons of a biological brain :

1- The first neurons called “input neurons” receive informations. 

2- multiple algorithms work together in these networks to process information.

3- The output data, “neural outputs” provide the result of the processing to the software embedded in the glasses.

In our case, the data that are administered to this neural network are the blinks and non-blinks, each characterized by their own curves.

This is called raw data (see photo above).

This technology embedded in the frames of our connected eyewear helps to limit risks and provide a better user experience. The advantage also lies in the ability of the eyewear to adjust to each individual and to learn on its own on a case-by-case basis. Technology continues to surprise us.