How Machine Learning Works
The process follows a specific flow: Training Examples lead to an Algorithm (a set of rules), which then uses Input Data to create a Prediction or Output. During training, the algorithm analyses data for patterns to create a mathematical model of the relationships within that data.
Three Types of Learning Paradigms (LP)
- Supervised: Inputs consist of labelled data. Examples include spam detection, image recognition, credit scoring, and medical diagnosis.
- Unsupervised: Inputs are not labelled. The model finds patterns between similar items to group them. Examples include customer segmentation, market basket analysis, anomaly detection (e.g., fraud), and image compression.
- Reinforcement: The algorithm learns from the environment through a system of rewards and penalties. Examples include game playing, robotics, and autonomous vehicles.