AI Blog

AI Blog

by Michele Laurelli

Bias Term

/ˈbaɪəs tɜːrm/
Concept
Definition

An additional learnable parameter in neural networks that allows shifting the activation function.

Bias terms enable neurons to activate even when all inputs are zero. Each neuron typically has one bias. They increase model flexibility and are crucial for fitting non-zero-centered data.

Examples

1

Linear transformation: y = Wx + b

2

Bias in each layer

3

Offset parameter

Michele Laurelli - AI Research & Engineering