Machine learning is a family of techniques that enable computers to learn patterns from data and make predictions or decisions without being explicitly programmed for each task. It is built on algorithms that adjust internal parameters to minimize error on training examples, then apply what they have learned to new data.

A core distinction is between supervised and unsupervised learning. In supervised learning, algorithms are trained on labeled examples, such as images tagged with object names or patient records labeled with disease outcomes. Tasks include classification, where inputs are assigned to categories, and regression, where continuous values are predicted. In unsupervised learning, data are unlabeled, and the goal is to uncover hidden structure, such as clusters or lower dimensional representations.

Neural networks and deep learning have driven many recent advances, especially in vision, language, and speech. These models stack many layers of simple computational units that gradually transform raw input into more abstract features. Training often relies on gradient descent and backpropagation to iteratively reduce prediction errors.

Machine learning research also explores reinforcement learning, in which agents learn to act by receiving feedback in the form of rewards and penalties. This framework has produced systems that master complex games and control tasks.

Across domains, applications include medical diagnosis, climate and environmental modeling, scientific discovery, and industrial optimization. At the same time, researchers study issues of interpretability, reliability, data quality, and bias. Understanding how models reach their decisions and ensuring they generalize well are central challenges for safely integrating machine learning into scientific and societal practice.