Module subspace_methods

Subspace methods for very high-dimensional optimization

This module implements various subspace methods that are effective for optimization problems with thousands to millions of variables. These methods work by restricting the optimization to lower-dimensional subspaces, making them computationally feasible for very large-scale problems.

Structs

SubspaceOptions

Options for subspace optimization methods

Enums

SubspaceMethod

Subspace method types

Functions

minimize_adaptive_subspace

Adaptive subspace method using gradient history

minimize_block_coordinate_descent

Block coordinate descent method

minimize_cyclical_coordinate_descent

Cyclical coordinate descent method

minimize_random_coordinate_descent

Random coordinate descent method

minimize_random_subspace

Random subspace gradient method

minimize_subspace

Minimize using subspace methods