Machine Learning is a family of methods where systems improve performance by learning patterns from data. Traditional software is programmed rule by rule; ML systems learn statistical relationships.

Types of machine learning

Dataset: examples used to train or evaluate a system.
Feature: a measurable property used by a model.
Algorithm: a procedure used to learn from data or make decisions.
Model: the learned structure produced after training.
Weights: internal numerical parameters adjusted during training.

Supervised learning: learns from labeled examples, such as emails marked spam/not spam.
Unsupervised learning: finds structure in data without labels, such as clustering similar customers.
Reinforcement learning: learns through feedback from actions and outcomes.

Neural networks are models inspired loosely by connected processing units. Deep learning uses many layers to learn complex representations. This is especially powerful for language, images, speech and large-scale pattern recognition.

The learner should remember one practical point: powerful models can capture subtle patterns, but they can also reproduce errors, bias or overconfidence if the data, objective or evaluation is weak.