Sigmoid Derivative
Domain: Math & Logic for AI
🎯 Motivation
The slope of sigmoid, used in backpropagation. Maxes at 0.25, which causes vanishing gradients in deep networks.
📋 Concrete Example
Sigmoid derivative at x=0 is 0.25 (steepest). At extremes it approaches 0, meaning gradients vanish.
🔗 Analogy
Like a hill that is flattest at the top and bottom - hard to slide down when there is no slope.