Skip to content
AI’s

training

Gradient Descent

An optimization algorithm that iteratively adjusts model parameters in the direction that most reduces the loss function. Gradient descent computes the gradient (slope) of the loss with respect to each parameter and updates them proportionally.

In practice

Imagine rolling a ball downhill to find the lowest valley; gradient descent similarly follows the steepest slope to minimize error.