Gradient descent

Gradient descent is an optimization method that repeatedly adjusts a model's parameters in the direction that most reduces the loss, taking small steps guided by the gradient.

Gradient descent sits in the Foundations part of the Agentik {OS} glossary, which defines the words used to build and run AI agent systems.