Definition
An early neural learning model that classifies inputs using a weighted sum and a threshold, and adjusts weights using misclassified examples.
Rosenblatt proposed the perceptron model in his 1958 paper. This entry focuses on the single-layer classification form that later became widely taught and used, rather than reducing all the structures of the original complete perceptron system to a single formula.
The single-layer model's expressive power is constrained by linear separability. This limitation should not be generalized to all multilayer neural networks.
Principles and impact
Problems solved
- Allowed some of a classifier's decision parameters to be learned from examples.
Principles
- Produces a classification from a weighted sum of inputs and corrects the weights using feedback from examples.
Improvements
Not recorded
Significance
- Provided an early approach combining neural computation with data-driven learning.
History and applications
People
Frank Rosenblatt
Organizations
Not recorded
Events
Not recorded
Applications
Linear classification、Teaching neural learning
Limitations
- Single-layer linear discrimination cannot directly express arbitrary nonlinear classification boundaries.
Upstream and downstream relations
Upstream
Sources: The Perceptron: A Probabilistic Model for Information Storage and Organization in the BrainLearning Representations by Back-propagating Errors
Downstream
Sources: The Perceptron: A Probabilistic Model for Information Storage and Organization in the BrainLearning Representations by Back-propagating Errors