Perceptron vs Multilayer Perceptron
People think a multilayer perceptron is just a big perceptron. It is more than that. One perceptron adds up its inputs and says yes or no, so it can only split data with a straight line. A multilayer perceptron stacks many of them in layers, with a bend in each one. The layers let it learn curved and tangled patterns that one perceptron never can.
Perceptron
An early machine from 1958. It learned to sort things by changing its weights.
The perceptron is one of the first machines that learned from examples. Frank Rosenblatt described it in 1958. It adds up its inputs, each with its own weight, and says yes if the total passes a set line. When it gets an answer wrong, it moves the weights a little. A single perceptron can only split data with a straight line, so it fails on many tasks.
Multilayer Perceptron (MLP)
The most plain kind of neural network. Each layer links to all of the next.
A multilayer perceptron is the most plain kind of neural network. It has an input layer, hidden layers and an output layer. Every neuron in one layer links to every neuron in the next. Data flows one way, from input to output, with no loops. It works well on rows of numbers, like a spreadsheet. Bigger networks, like transformers, still have MLP parts inside them.
In one line
- Perceptron
- An early machine from 1958. It learned to sort things by changing its weights.
- Multilayer Perceptron (MLP)
- The most plain kind of neural network. Each layer links to all of the next.