AI Blog

AI Blog

by Michele Laurelli

Momentum

/moʊˈmɛntəm/
Technique
Definition

An optimization technique that accelerates gradient descent by accumulating a velocity vector in directions of persistent reduction in the loss.

Momentum helps navigate ravines and accelerates convergence. It adds a fraction of the previous update vector to the current one, smoothing optimization trajectory and reducing oscillations.

Examples

1

SGD with momentum 0.9

2

Nesterov momentum

3

Accelerating convergence

Michele Laurelli - AI Research & Engineering