torch_numopt.algorithms.gauss_newton

Gauss-Newton optimization algorithms for least-squares problems.

The Gauss-Newton method approximates the Hessian as JᵀJ, where J is the Jacobian of the residual vector. This module provides three variants: a vanilla version with a fixed learning rate, a line-search version, and a trust-region version. The block-diagonal approximation is available for memory efficiency.

Classes

GaussNewton(params[, lr_init, lr_method, ...])	Gauss-Newton optimizer (no line search or trust region).
GaussNewtonLS(params[, lr_init, lr_method, ...])	Gauss-Newton optimizer with line search for step-length selection.
GaussNewtonTR(params[, lr_init, ...])	Gauss-Newton optimizer with a trust-region subproblem solver.