gradient-descent

Gradient descent is a gradient-based local optimization method. This is probably the simplest method in this category.

#include <mathtoolbox/gradient-descent.hpp>

Internal Dependencies

Math and Algorithm

This implementation uses backtracking-line-search to find an appropriate step size. The initial step size needs to be specified as default_alpha.

Stopping Criteria

(TODO)

Bound Conditions

This implementation supports simple lower/upper bound conditions.

Useful Resources