digital drawing of a face and binary coding

Reading time: 5 min

Your favorite music app suggests songs that match your tastes; the platform where you watch TV shows and films makes suggestions that you might like; and your favourite online clothing shop recommends clothes and accessories that match your style. How can we explain this? Are machines capable of reading your mind? They don't go that far yet. Behind these recommendations is Machine Learning, a branch of artificial intelligence that allows computers to learn autonomously and perform actions without needing to be programmed. 

What is Machine Learning?

Machine Learning is a branch of artificial intelligence that enables machines to learn autonomously without needing to be specifically programmed to do so.

It is not a new concept. The history of machine learning started with an IBM employee, Arthur Samuel, who coined the term in 1959. Samuel was a pioneer in the field of computer games and achieved one of the most famous machine learning stories to date. He managed to get an IBM 701 computer to beat a human at checkers. In the 1990s, especially with the development of the Internet and the increase in the amount of data that could be accessed, Machine Learning experienced a remarkable growth. Among other achievements, a computer called Deep Blue was able to beat Gary Kasparov himself at chess in 1996. 

Today, it is used in a wide variety of fields, such as facial recognition through software, spam detection by email providers, and voice recognition, among others.

Differences between Machine Learning and Deep Learning

Machine Learning and Deep Learning are both branches of artificial intelligence. Although both concepts are sometimes used interchangeably, Deep Learning is a subtype of Machine Learning. 

Very simply, they are learning algorithms (regression or classification) whose objective is to obtain a result dependent on the input variables (data). Conceptually it would be something similar to: if we have "a" and "b", then the result is "x", but if we have "a" and "b" plus "c", then the result is "y".

The fundamental difference between the two concepts lies in the way the respective algorithms learn

Methods and types of Machine Learning

There are three main methods of Machine Learning, based on the algorithms they employ (i.e., sets of operations that are systematically executed and perform calculations to find the solution to a problem): 

drawing of a brain

  • Supervised learning: algorithms deduce information from data that have been assigned values. This data is used to train a model, while a second set of test data is used to determine how effective the created model is. An example would be the calculation of the price of a house from its characteristics. It has similarities with a new method called "reinforcement learning". In this type, the system learns from the mistakes it makes until it finds the optimal way to perform a task. 
  • Unsupervised learning: Training data is used without labelling. Algorithms detect clusters of data or hidden patterns without requiring human intervention. They are very useful for discovering similarities and differences in information, so they can be applied to processes such as customer segmentation or image recognition.
  • Semi-supervised learning: training data with and without labels is used; typically the labelled data set is much smaller. They are used, for example, to analyze conversations in a call center to easily deduce the characteristics of the callers and their mood, among other aspects.

Repsol and Machine Learning

a hand on a keyboard superimposed with icons from apps

At Repsol, we are firmly committed to the digital transformation. That is why we have committed to the development of ARiA (Advanced Repsol Intelligence & Analytics), a cloud-based data and analytics platform. 

This platform, created in collaboration with Accenture, includes Machine Learning models, configuration alternatives in the Microsoft Azure cloud, tools for data governance from its development laboratory, and solutions for data ingestion. 

It enables employees to make data-driven decisions with self-discovery capabilities, regardless of their analytical skills.