In a study conducted by MIT’s Technology Review of more than 16,000 Artificial Intelligence research papers, reinforced learning emerged as a new trend which has seen a rise in recent years, very much in the same way as deep learning. Reinforced learning is actually one of the three different types of machine learning that, unlike supervised and unsupervised learning, emphasizes on rewards and punishment to train models.
Vaclav Vincalek, a tech entrepreneur and partner with Future Infinitive claims that the concept of reinforced learning has been around for a very long time, something for which cats should be thanked for.
Psychologist Edward Thorndike experimented sometime in 1898, in which he placed a cat in a box that could be opened only if the lever in the box is pulled. He put a scrap of fish outside, which motivated the cat to escape. In each turn, the cat would take some considerable amount of time to figure out its way. With each successive trial, it took the cat less time to get out of the box. Thus, this proved that action is more likely to be repeated if it has a pleasant consequence and less likely to be repeated if it has a negative consequence.
The concept of reinforced learning is now used in many areas such as robotics, telecommunications, and business strategy planning. It has also been used by companies such as Mobileye, Google and Uber in AI driving simulations, where virtual cars are required to complete a course over and over again. It has also been said that the application of reinforced learning will be soon used for selecting online ads and automating financial trading.