Kom jij in aanmerking voor subsidies of rentevrije leningen? - Ontdek het hier
Terug naar alle artikelen

23 juli 2020

Wat is machinaal leren?

Ironhack - Changing The Future of Tech Education

Data Analytics

Alle Cursussen

In recent decades, the technology industry has seen the most growth in the field of artificial intelligence and, more specifically, machine learning. In a world where data has become a highly valued commodity, Machine Learning has gained enormous relevance in today's tech ecosystem.

What is the purpose of machine learning?

The main goal of machine learning is to allow machines to learn completely on their own, without anyone having to perfect their algorithms. The aim is that - just like the human mind - they can improve their own processes, so that they can perform the tasks entrusted to them with an increasing degree of precision. For machine learning to really work, it is necessary to provide the machine with certain information, either by providing the necessary information through files loaded with a multitude of data, or by letting the machine collect data through its own observations and even interact with the real world. 

This kind of learning, through the collection and interpretation of data, has allowed them to move from relatively simple tasks to more complex tasks. Initially, they were willing to filter emails or perform other day-to-day actions, but over time they have gained more computing power and today they are capable of analyzing much more complex tasks. Now they are able to analyze millions of medical diagnostic results, which, among other benefits, may enable us to detect cancer more reliably; it can predict traffic patterns, allowing us to plan routes in a timely manner; they can even execute architecture projects in real time, simply by recognizing a particular area. 


What types of algorithms are there in machine learning?

In artificial intelligence (or AI), there are three primary types of machine learning algorithms that are used depending on how the machine is trained and instructed to improve task performance. The end goal is for the machine to perform actions in an increasingly optimized way by refining patterns and behavior through continuous learning. The three main types of learning algorithms are: Supervised, Unsupervised, and Reinforcement. To decide what type of machine learning is needed, it's fundamental to know exactly what the goal is that we want to achieve by programming that artificial intellect.

Machine Learning Algorithms - Towards Data Science

Supervised Learning Algorithms

This type of machine learning is all about providing machines with information upfront so that they have first examples and can expand their knowledge over time. It is usually done through labels, which means that when we program the machines, we pass them correctly labeled elements so that they can label new elements later without human intervention. For example, we can pass the machine pictures of cars, buildings, road signs or anything else relevant to our job, then we tell what each item is and how we want it to be interpreted. With these first examples, the machine generates its own knowledge base so that it can continue to give labels when recognizing a car, a building or a traffic sign.

In this type of machine learning, the machines are not limited to training images, but can use different data types. Provided with sounds or calligraphy data sets, they can learn to recognize voices or detect written patterns and associate them with a particular person. The possibilities arise entirely from the initial data supplied to the machine.

Unsupervised Learning Algorithms

In this case, the machine will not be provided with any kind of previously labeled information about what it should recognize, meaning it will not have an existing knowledge base. Instead, it gets data about the characteristics of the case it needs to identify and should learn to recognize those characteristics itself. Essentially, this type of learning algorithm requires the machine to develop its own knowledge base based on a limited data set.

This is ultimately the closest thing to the way the human mind learns and develops. The machine learns to analyze groups using a method called clustering. This is nothing more than grouping the elements according to a set of characteristics they have in common.

Reinforcement Learning Algorithms

This form of machine learning is applied when systems are intended to learn from past experience. In this case, when people program the algorithm, they define what the final result should be without specifying how best to achieve it. The machine is thus responsible for discovering how to achieve its goal. The machine is responsible for running a series of tests in which it makes mistakes and successes, learns from the latter and ignores the other actions that led to failure. Basically, it detects patterns of success that it repeats over and over to become more and more efficient.

Autonomous cars are a good example of this kind of learning algorithm. Their job is very clear: to take passengers to their intended destination. As the cars travel more and more, they discover better routes by identifying shortcuts, roads with fewer traffic lights and more. This allows them to optimize their journeys and thus do their work more efficiently.

Machine learning versus Deep Learning

Many consider the two fields to be interchangeable, but actually Machine Learning and Deep Learning are very different disciplines. Broadly speaking, the latter is a kind of subset of machine learning, as its technical concept and practical application are nothing more than a form of machine learning. This is why some people often confuse or use the two terms interchangeably.

Essentially, what Deep Learning does is structure algorithms to form an artificial neural network, allowing it to both learn and make autonomous decisions. This is exactly where the main difference between the two technologies lies. For a machine programmed with some kind of Machine Learning, a human has to correct errors of the machine by modifying its configuration to prevent it from failing again. However, a Deep Learning model can determine for itself whether its task has been completed successfully or not, by using its own neural network.

Ultimately, there are almost endless possibilities that come from the exponential growth of Machine Learning. That's why technology companies are increasingly looking for candidates with extensive knowledge of Data Analytics. For those passionate about AI and machine learning, the Ironhack Data Analytics Bootcamp serves as a gateway to progress in this fascinating technological world.

Vergelijkbare Artikelen

Aanbevolen voor jou

Klaar om mee te doen?

Meer dan 10,000 carrièreveranderaars en ondernemers lanceerden hun carrière in de tech industrie met Ironhack's bootcamps. Start uw nieuwe carrière reis en sluit u aan bij de tech revolutie!

Ik accepteer de Privacybeleid en de Gebruiksvoorwaarden