Machine Learning in the Essence

The Frontier of Machine Learning / Machine Learning in the Essence

You open the newspaper page and see an article written entirely by a robot, scary is that?, don’t worry we are in the age of Artificial Intelligence (AI) and the article was an initiative of The Guardian Newspaper in September 2020 with the aim to show the power of the algorithms used by AI and their impact on human life. A great transformation is taking place in many areas, whether in medicine, engineering, business, finance. But what really is Machine Learning?

The highly regarded journal Nature provides the following definition of “Machine learning is the ability of a machine to improve its performance based on past results. Machine learning methods allow computers to learn without being explicitly programmed and have multiple applications, for example in improving data mining algorithms.”

We can observe that Artificial Intelligence is a subject widely publicized by movies, series etc. However, in recent years terms such as (ML, Machine Learning) and (DL, Deep Learning) have gained the attention of numerous periodicals and journalistic articles. However, they do not mean the same thing, they are part of the same universe and are directly related.

AI is not something new, its theoretical basis goes back to the 1950s. As well as the fundamentals of Machine Learning and Deep Learning, which have been published for decades.
The term Artificial Intelligence appeared in the year of 1956, in a conference in the USA.
In the beginning, an AI-based system basically operated with a set of rules previously defined by a human being. Such rules sought to describe situations in which, if a pattern was identified, a problem could be solved and, as an example, we have the credit card fraud detection system.

Machine learning is a subarea of ​​AI, which at its most basic aspect is the practice of using algorithms to analyze data, learn from them, extract patterns, and then make a determination or prediction about something in the world, without being explicitly programmed for that.

A limitation of Machine Learning is the type of data it can receive, which can only be structured. This means that these algorithms can only process information that is arranged as in a traditional database, and cannot work with images and sounds, called unstructured data. And it is precisely at this point that Deep Learning and all its revolution comes in. algorithms are triggering.

ML is a set of algorithms, each of which serves a type of task. Such algorithms are classified into 4 large groups, based on the way learning works internally.

The first group is Supervised Learning
The second group is Unsupervised Learning.
The third group is called Learning by Reinforcement
And the last one is called Deep Learning

Since the focus of the machine learning field is “learning”, there are 14 types that you can find as a professional divided into subfields of learning and which include statistical inferences and learning techniques.

However, algorithms that generate trainable models in ML are not as smart as most people think. They need to learn and to learn they need data. A trainable model must also be tested and evaluated for it to be viable.
Shawn Ennis in his article demonstrates how workflow for machine learning is, as shown in the figure below:

Today AI is used to design evidence-based treatment plans for cancer patients, instantly analyze results from medical tests to escalate to the appropriate specialist immediately, and conduct scientific research for drug discovery.