Weighted Inputs and Bias

One unit combines evidence

A perceptron is a simple artificial neuron. It multiplies each input by a learned weight, adds the products, adds a learned bias, and passes the result through a decision function. A positive weight supports activation, a negative weight opposes it, and magnitude expresses influence under the feature scale.

inputs:  error_rate=2, queue_depth=3
weights:             0.8,           0.4
bias: -2
score = 2×0.8 + 3×0.4 - 2 = 0.8
step(score) -> 1

The bias shifts the boundary independently of inputs, like the intercept in linear regression. With two inputs, the perceptron creates a straight decision boundary; in more dimensions it creates a flat hyperplane.

Analogy: Each input is a witness, each weight determines how much that witness counts, and the bias is the starting presumption. Add the testimony and compare it with the decision threshold.

Note: Modern neural networks use smoother activation functions and gradient-based training, but this weighted-sum skeleton remains inside every dense layer.