# Stochastic Gradient Descent **Domain:** Optimization / Machine Learning **Doc Type:** Canonical Method Node **Maturity:** Foundational ## Definition **Stochastic gradient descent** updates model parameters using gradients estimated from individual examples or subsets of data rather than from the complete dataset at every step. ## Historical Context [[wiki/Shun-Ichi Amari|Shun-Ichi Amari's]] 1967 probabilistic descent method is interpreted by [[wiki/Jürgen Schmidhuber|Jürgen Schmidhuber]] as an early stochastic-gradient procedure applicable to multilayer training. The attribution remains distinct from the paper's own vocabulary. ## Technical Context Stochastic estimates reduce the cost of each update and introduce variation into movement through an [[wiki/Optimization Landscape|optimization landscape]]. ## See Also [[wiki/Gradient Descent|Gradient Descent]], [[wiki/Local Optimum|Local Optimum]], [[wiki/Credit Assignment|Credit Assignment]]