Perceptron

Also known as · neuron

The basic unit of a neural network: a weighted sum of its inputs plus a Bias, passed through an Activation Function to give one output.

When to use

Exam questions ask you to compute one by hand: multiply each input by its Weight, add the bias to get zz, then apply ReLU or sigmoid. Alone it only learns a linear decision boundary (like Logistic Regression), so it cannot capture an interaction such as rain mattering more in the evening.

See DS Lec 8.

Appears in

No references yet.