Show in graph
AI

Data → Machine Learning & AI

Gradient Descent

An iterative optimization method that updates parameters in the direction that reduces a loss function.

Motivation

Gradient descent solves the problem of improving parameters by repeatedly moving in the direction that reduces loss.

Where it fits

It belongs to optimization and model training.

Mental model

Imagine standing on a landscape and walking downhill. The gradient tells you which direction is uphill, so you step the other way.

Important details

Important details include learning rate, local minima, stochastic batches, and convergence behavior.

Common mistakes

  • Using a learning rate that is too high or too low.
  • Assuming optimization success means the model generalizes.