Machine learning is a computer process that is frequently used in the implementation of artificial intelligence. Machine learning is a concept that is becoming more and more important in computing, but this term is often unknown to the general public. How does a machine learn? The answer in this article.
Definition of machine learning
Machine learning is a concept that works closely with artificial intelligence. This term reflects an approach in which a machine will be able to learn on its own, to be able to provide an answer following complex operations.
The main objective of this concept is to enable a machine, or computer, to provide automatic solutions based on data previously received. This machine learning, also known as statistical learning, involves the ability for a machine to process a colossal amount of data in a short time and to be able to learn from choices made previously or not.
This type of approach is a revolution in the field of artificial intelligence and opens the doors to an immeasurable number of possibilities. Nevertheless, even as machine learning begins to evolve considerably and expand in the artificial intelligence market, there are still technological limitations to the massive use of this concept.
Different learning processes
Machine learning is therefore a concept that allows a machine to make strategic choices based on data and previously recorded information. To do this, the machine or computer, goes through a complex computer program or algorithm, allowing it to learn automatically.
There are now two major types that are:
- Supervised learning or discriminatory analysis;
- Unsupervised learning or classification.
Supervised learning or discriminatory analysis
Supervised learning is done using data that has already been catalogued and classified. The next step will be to "sort" the new data introduced, to find out which class it belongs to.
The machine's response will be in relation to models previously introduced into the system. In this case, the learning takes place in two stages. The first is to model catalogued data, or in other words, to provide responses based on the value of the data.
The second step is to base the answer on new data based on the previously defined models, so that we can catalogue this new data and take it as a model for future data.
Unsupervised learning or classification
Unsupervised learning is by no means based on pre-defined elements. It is the machine or computer that will categorize the data received by itself by looking for certain similarities. This type of method therefore allows for real machine learning, where the machine is not supervised by man and where it carries out associations on its own, compared to raw data not yet categorized.
The difference between supervised and unsupervised learning
The big difference between supervised learning and unsupervised learning is data labelling.
In the case of the first type of learning, data is already classified and assimilated by the machine, which will then "simply" compare the new data with the pre-existing models, to provide a more or less pre-established response based on the value of the new data.
In the second case, the machine gets all the raw data and will itself try to find relationships within that data set, to provide a specific and unspecified response. This response can go beyond human understanding of a situation.
Here's an example to illustrate the difference between these two learnings:
- Supervised learning: as previously explained, in supervised learning, the system already has pre-established models and data. Let's say these models are triangles and circles. When new data appears, here a green triangle, it is processed by the machine and categorized into the triangles. In this case, the machine only takes into account the form since the established models only involve the form;
- Unsupervised learning: in the case of unsupervised learning, raw data is therefore subject to processing by the machine, which will classify it according to the criteria it itself will consider. It will therefore be able to classify information by form, colour, shape and colour, thus adding a much larger number of parameters than could have been thought of by man.
It is very important to add that some machine learning systems involve both types of learning, to achieve more precise results and thus avoid certain deviations that could result from unsupervised learning.
Differentiating machine learning and artificial intelligence
Artificial intelligence and machine learning are both part of computer science, however, even though these two technologies are closely associated with each other, they are still distinct, and many people use both terms as synonyms, wrongly.
Here is a comparative table to differentiate between artificial intelligence and machine learning.
|Simulates human behavior||Allows you to automatically learn data without explicit programming|
|The goal is to create a computer system just as intelligent as humans to solve complex problems||The goal is to allow machines to learn data so that they can produce accurate results.|
|Create intelligent systems to perform any task as a human being.||Teach machines from data to enable them to perform a particular task and give a specific result.|
|Machine learning and in-depth learning are the two largest sub-sets of artificial intelligence.||Deep learning is the great subset of the Learning machine.|
|Wide range of possibilities||Opportunities are limited|
|Example of artificial intelligence: Siri||Example of machine learning: suggestion of self-tag, of friends on Facebook.|
|Artificial intelligence learns, reasons and self-corrects.||Machine learning learns and self-corrects with new data but does not reason.|
Examples of machine learning
E-commerce and Amazon
Amazon demonstrates the place of machine learning in e-commerce, with their product recommendation system, based on the research you've done beforehand and predicting your future needs.
Health and IBM
IBM has developed a machine learning algorithm that can synthesize texts taken by doctors using a technique called Natural Language Processing and then establish criteria for diagnosing heart failure, just as a cardiologist would.
Music and Machine Learning for Drummers
Peter Sobot has developed an app called Machine Learning for Drummers that, as the name suggests, helps drummers with an application using machine learning. This application analyzes samples of percussion audio and is able in 87% of cases to determine what type of percussion is causing the noise (Kick drum, snare drum or another type).
Machine Learning and YouTube
YouTube uses machine learning a lot in its algorithms. Whether it's for suggestions compared to what you've watched before, to censor, or demonetize certain videos.
Machine Learning is a technology that is part of the field of artificial intelligence and allows a machine to learn, based on a certain number of data. It's not always easy to use your data, which is why Ryax supports you in exploiting it.
The Ryax Team.