Perceptron Learning Algorithm Demo - MATLAB

less than 1 minute read

Code available here

This is the MatLab implementation of the popular basic classification algorithm known as Perceptron Learning Algorithm(PLA)

The purpose is to show how PLA converges to its final hypothesis

For a linearly seperable dataset PLA converges to one final hypothesis with perfectly classified classes
For a linearly inseperable dataset PLA does not converge to any hypothesis and we choose the hypothesis of the last iteration as out final hypothesis

Running the code

In the main.m file change the ‘type’ variable to 1 to create linearly seperable points
Change ‘numpoints’ variable to the number of datapoints to be taken
In the MatLab command line run the command
```
main.m
```
Linearly inseperable
In the main.m file change the ‘type’ variable to 0 to create linearly inseperable points
Change ‘numpoints’ variable to the number of datapoints to be taken
In the MatLab command line run the command
```
main.m
```
Result

Linearly seperable

The PLA converges to a hypothesis in case of linearly seperable. The final hypothesis division is represented by the green line.

alt

The PLA does not converge in case of linearly inseparable, the magenta line represents the final hypothesis achieved after 5000 steps