# Perceptron **Entity class:** Learning algorithm and neural model **Domain:** Machine learning / neural networks **Doc Type:** Developed Method Node **Maturity:** Foundational ## Definition The **perceptron** is a trainable linear classifier introduced by [[wiki/Frank Rosenblatt|Frank Rosenblatt]] in 1958. It combines weighted inputs, a threshold decision, and an error-driven update rule. ## Historical Role The perceptron converted the formal-neuron lineage into a system that could adjust from examples. Rosenblatt's Mark I Perceptron embodied the method in hardware. Later multilayer neural networks expanded beyond its linear decision boundary, while [[wiki/Perceptron Branch Predictor|perceptron branch predictors]] carried the core classifier into commercial processors. ## Evidence Ledger - **Established:** Rosenblatt defined the perceptron and its learning procedure in 1958. - **Established:** The Mark I implemented trainable classification in hardware. - **Strongly indicated:** The perceptron is a transmitted objective between formal neural models, learning machines, and later connectionism. - **Unresolved:** The strength of direct lineage into any particular modern architecture; publication citations and designer testimony decide individual claims. ## Sources / Provenance - Frank Rosenblatt, [“The Perceptron” (1958)](https://doi.org/10.1037/h0042519). - [[articles/A History of Machine Intelligence|A History of Machine Intelligence]]. ## Relationships - **Developed by:** [[wiki/Frank Rosenblatt|Frank Rosenblatt]]. - **Builds on:** [[wiki/Warren McCulloch|Warren McCulloch]] and [[wiki/Walter Pitts|Walter Pitts]]. - **Implemented in:** the Mark I Perceptron and [[wiki/Perceptron Branch Predictor|Perceptron Branch Predictor]].