Local Minimum
Domain: Math & Logic for AI
🎯 Motivation
A low point that is not the lowest overall. Gradient descent can get stuck here.
📋 Concrete Example
On a bumpy loss surface, the gradient might lead to a small valley that is not the deepest point.
🔗 Analogy
Like resting in a small ditch on a mountainside - you are low locally, but not at the valley floor.