The Perceptron: A Single Artificial Neuron
What you'll learn: How a perceptron combines weighted inputs with a bias term and a step activation function to create the simplest building block of neural networks.
The Core Idea
A perceptron is the most basic artificial neuron, designed as a linear classifier. Think of it as a decision-maker that takes multiple inputs, weighs their importance, and produces a simple yes/no output.
Here's how it works:
- Weighted Inputs: Each input feature gets multiplied by a learned weight (how important is this feature?)
- Sum with Bias: Add all weighted inputs together, plus a bias term (a threshold adjustment)
- Step Activation: If the sum exceeds zero, output 1; otherwise, output 0
Mathematical Form
The perceptron computes:
z = w₁x₁ + w₂x₂ + ... + wₙxₙ + b
output = 1 if z ≥ 0, else 0
Where w are weights, x are inputs, and b is the bias.
An Analogy
Imagine a hiring decision: you score candidates on skills (inputs), weight each skill by importance (weights), add a baseline preference (bias), and hire if the total score passes a threshold (step function). The perceptron does exactly this for classification.
Why "Linear Classifier"?
The perceptron draws a straight line (or hyperplane in higher dimensions) to separate two classes. It can only solve linearly separable problems—those where a straight boundary can perfectly split the data.
Key Takeaway: The perceptron is a linear binary classifier that combines weighted inputs and a bias, then applies a step activation function to produce discrete outputs—serving as the historical foundation for modern neural networks.