Machine Learning or Machine Learning is a field of study that teaches computers to analyze and classify data patterns to make predictions. The concept is part of the Computer Science area and combines model recognition and Artificial Intelligence applications. It can be summed up Machine Learning simulating a natural process for human beings: learning from experience.
Software and PCs have the potential to process large volumes of data, as well as resources to identify and separate models and patterns. Machine Learning is a programming of systems to assimilate data and classify complex information, with characterization of the learning, to then present forecasts and estimates.
All about artificial intelligence: 10 facts you need to know
Android P use machine learning to identify which apps the user will need – and thereby save up to 30% more battery Photo: Nicolly Vimercate / dnetc
The area has more and more everyday applications with the sophistication of mobile apps. Movie suggestions on Netflix and music on Spotify, as well as alternative routes on Google Maps and Waze characterize the private jobs of this technology compared to tips on the buying behavior of customers and public surveillance cameras with facial recognition.
With Machine Learning, it is possible to increase the human capacity to solve problems and anticipate risks, based on the results raised by the programs. This applies to issues related to Big Data, that is, the large set of stored data, in which Machine Learning has been the key technique for solving demands. Uses range from medical diagnostics, weather forecasts and identification of climate change to analyzes and deductions on the stock market.
Spotify uses Machine Learning to suggest music more suited to the user's taste Photo: Reproduo / Daniel Ribeiro
Artificial Intelligence vs Machine Learning
Artificial Intelligence is an area of study developed in the 1950s, which proposes computer programs with levels of communication and successful responses, in simulation of intellectual interactions. In addition, the topic still aims to create knowledge models that provide automatic responses based on data analysis and user observation.
Machine Learning is a subfield of Artificial Intelligence Photo: Reproduo / Creative Commons
Machine Learning, as mentioned earlier, can be considered a subfield of science in which Artificial Intelligence is. However, its specificity is based on the knowledge acquired by computers, which makes it possible to anticipate facts and behaviors according to user patterns previously identified.
Thus, Machine Learning can be considered as an important part of Artificial Intelligence, as it improves the experience acquired by the computer. In this way, programs can be improved faster with techniques ranging from face recognition or handwriting, to thematic identification of photos in an image gallery or the simultaneous translation of plates and menus with the cell phone camera.
Machine Learning programs, such as Google Images, learn to classify data references and group by themes Photo: Reproduo / Daniel Ribeiro
These innovations are a direct consequence of improving the assimilation and classification of data by Machine Learning. Such technological benefits are directly linked to the processing of large volumes of data by computers, instead of programmed instructions, line by line, in programming language.
How does Machine Learning work?
The Machine Learning program is divided into Supervised Learning and Unsupervised Learning. The first establishes known data entry models and prediction output, while the second identifies hidden patterns and structures in data entry.
Machine Learning is subdivided into Supervised Learning and Unsupervised Learning Photo: Reproduo / Daniel Ribeiro
Supervised Learning determines a learning algorithm from a set of known data to classify the information. In addition, in parallel to this categorization system, the system can still recall previous data entries to make predictions and deductions based on the groups of information already incorporated.
Unsupervised Learning refers to programming that finds hidden patterns or special structures in the data. Also called "Clustering", this specification allows the construction of estimates for complex information and without records in the system, since the crossing of the most successful responses constitutes the condition of "knowledge" of the machine.
Gmail's Smart Response is a Machine Learning feature Photo: Reproduo / Daniel Ribeiro
Thus, Machine Learning has internal procedures that characterize the processing of elements. For example, Supervised Learning ensures that email recognizes spam messages, while Unsupervised Learning makes it possible to use the new Smart Reply in Gmail.
Examples of Machine Learning in everyday life
Many software used daily have Machine Learning technology. Continuous improvements to e-mail apps, GPS navigation and even browsers offer improvements adapted to the user, according to customs, the way of writing and the browsing history.
Google Maps' route suggestions take into account information on the route and traffic according to the schedule Photo: Reproduo / Daniel Ribeiro
The innovations influence the creation of "Bubble Filters", with information selected according to the user's tastes. There are also programs with noble and selfless goals, such as TensorFlow, Google Translate's open Machine Learning platform. It is used in research projects to track animals at risk of extinction or to diagnose eye diseases in diabetics.
Other apps have more particular features, like Google Play Music, with recommendations for playing according to the weather or time of day. Or Google Maps which, in addition to recognizing street names and addresses of billions of Street View images, takes into account traffic routes and the availability of parking in city regions, depending on the time of day.
Slice was one of the new tools with Machine Learning presented at Google I / O Photo: Reproduo / Google
Google I / O, an annual event with news from the company, presented a series of innovations to improve the personalization and optimization of the use of smartphones with Android. Another example of using Machine Learning are bot projects with messengers and social networks, to assist in checking data and guiding users on how to surf the Internet safely and responsibly, in order to prevent the spread of false news. . In January, Facebook announced support for projects to combat fake news in Brazil with the use of this technology.
Questions about Machine Learning or Artificial Intelligence? Ask in dnetc's Forum.