One of the most basic of ML algorithms - but it has a cool name!
Method for learning a linear classifier: h(x;θ,θ0)=sign(θT+θ0)
- Can also be used to learn non-linear classifiers through methods like Polynomial Basis
The basic idea is:
- If a prediction is correct (sign is positive), make no changes
- If a prediction is incorrect (negative), edit θ and θ0
Python implementation:
Variation: Averaged Perceptron