Back to timeline

The Perceptron

Frank Rosenblatt publishes the Perceptron, an early trainable model that shaped the neural-network lineage of modern AI.

Research

What Happened

In November 1958, Frank Rosenblatt published “The perceptron: a probabilistic model for information storage and organization in the brain” in Psychological Review. The perceptron was presented as a learnable system for classification-like tasks.

Why It Matters

The perceptron became an iconic early example of a trainable machine learning model and helped establish learning-based approaches as a serious direction, foreshadowing the later resurgence and scaling of neural networks.

Technical Details

The perceptron can be described as a linear model with adaptive weights trained from data, an ancestor of later multilayer neural networks and gradient-based learning systems.