As the name suggests — Machine Learning is the ability to learn from any given data. Just like us mortal humans, the computers are competent to learn and perform tasks without being explicitly programmed to do so. The term machine learning was coined by Arthur Samuel in 1959, who was a pioneer in the field of computer gaming and artificial intelligence.
To understand it better, let us consider a scenario where you are playing your FIFA for the first time. Now obviously in your first ever match you are going to find some difficulties in control and gameplay. But as you progress you will learn and understand various information like abilities of every player in your team, the ideal position to take a shot, and passing techniques thereby empowering you to win your first ever league cup. So, during every match we were learning something new and trying to imply it into our next match to improve our performance. In short, we were programmed to learn from our experiences.
While watching your favorite series on Netflix, you might have noticed that the Netflix itself recommends you some of its shows which they think you might be interested in. This is the due to the courtesy of “Recommendation Engine” in Machine Learning. The engine had studied your preferences based on the shows you had watched, and therefore it predicted the shows you would like.
Classification of Machine Learning based on approaches
- Supervised Learning - This is the type of Machine Learning algorithm where machines are trained using well "labelled" training data. This is similar to that of human learning under the supervision of a teacher. In the supervised method, both input and correct output data is provided to the model.
- Unsupervised Learning - Contradictory to the Supervised Learning, the models are trained using unlabeled dataset and are allowed to act on that data without any supervision. The model itself has to find patterns from the given data. For instance let us consider a dataset consisting of images of cats and dogs, the model doesn't know either of them nor it has any previous idea about their features. In this case, it will use the Unsupervised Learning algorithm and will perform this task by clustering the image dataset into the groups according to similarities between images.
- Reinforcement Learning - When any dataset is provided which has examples that lack labels, just as in unsupervised learning. But they are accompanied by positive or negative feedback. Reinforcement learning involves taking decisions sequentially. Simply put, we can say that the output depends on the status of the current input and the next input depends on the output of the previous input. To have a glimpse of how exactly the technique works, you can take a look at Google DeepMind's reinforcement learning program, which plays old Atari video games. As you watch the video, note how the program is initially clumsy and unqualified, but gradually improves with training till it becomes a champion.