Learning is a human behavior. It gives birth to intelligence. When this intelligent behavior is simulated in machines, it’s known as Artificial Intelligence or AI. It includes but is not limited to acquisition of information, reaching conclusions, self-correction etc.
Machine Learning (ML) is generally seen as subset of artificial intelligence where machines learn from patterns and inferences in data to create/improve any statistical models. Based on its input, supporting environment and output for learning, ML can be categorized in different segments.
The different types of Machine Learning algorithms include Supervised Learning, Unsupervised Learning, and Reinforcement Learning. Let’s understand these terms one by one:
In Supervised Learning we have input data and reference output (working as supervisor) to learn from data or to find pattern in data. For example while we try to teach a child “how to identify character” we have an reference image/character to show him. The most common supervised learning algorithms are classification and regression.
In Unsupervised learning, we have no reference output (labeled data) to learn from data. The only thing we have is the data and we try to find patterns in data without any external reference or guidance. The unsupervised learning algorithms are used for clustering and anomaly detection in a large datasets.
The reinforcement Learning can be considered as special type of Supervised learning where we don’t have any direct output reference (labeled data) for learning but we do have punishment/reward for every learning outcome from our data which reinforces our learning behavior.