Although it is currently boosted by new technologies and new uses, machine learning is not a new field of study. We find a first mention in 1959, due to Arthur Samuel, one of the pioneers of artificial intelligence, who defines machine learning as the field of study aimed at giving the ability to a machine to learn without being explicitly programmed. In 1997, Tom Mitchell of Carnegie Mellon University offered a more precise definition:
"A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E".
A robot learning to drive? Let's briefly illustrate what machine learning can do with a simple case, probably closer to your everyday life: an anti-spam filter.
At first, imagine the "Machine" (your email service) "analyzing" which of your incoming emails will be classified as spam. Thanks to this "learning period", the machine deduces some major classification criteria. For example, the probability of the machine classifying an email as spam increases if the email contains terms such as "money". This can also happen if the sender of the email is not in your address book. Machine learning models either use supervised learning via labeled data sets to train algorithms for a particular output, or unsupervised learning via unlabeled data, allowing the algorithm to respond to new information without supervision.
Machine learning is a modern science for discovering patterns and making predictions from data based on statistics, data mining, pattern recognition, and predictive analysis. The first algorithms were created in the late 1950s.
Traditional analytical tools are not powerful enough to fully exploit the value of Big Data. The data volume is too large for comprehensive analysis, and the correlations and relationships between them are too important for analysts to test all the assumptions to derive any real value.
Additional Machine Learning Resources: