Skip to main content
Ctrl+K

torch_numopt 1.0.2 documentation

  • Quickstart
  • API reference
  • Module reference
  • Quickstart
  • API reference
  • Module reference

Section Navigation

  • torch_numopt.algorithms
    • torch_numopt.algorithms.adahessian
    • torch_numopt.algorithms.conjugate_gradient
    • torch_numopt.algorithms.gauss_newton
    • torch_numopt.algorithms.gradient_descent
    • torch_numopt.algorithms.hutchinson_newton
    • torch_numopt.algorithms.lbfgs
    • torch_numopt.algorithms.levenberg_marquardt
    • torch_numopt.algorithms.newton
  • torch_numopt.curvature
    • torch_numopt.curvature.exact_block_hessian
    • torch_numopt.curvature.exact_hessian
    • torch_numopt.curvature.gauss_newton_approximation
    • torch_numopt.curvature.gauss_newton_block_approximation
    • torch_numopt.curvature.hutchinson_diagonal_approximation
    • torch_numopt.curvature.naive_identity
  • torch_numopt.utils
    • torch_numopt.utils.param_operations
    • torch_numopt.utils.stability
    • torch_numopt.utils.utils
  • torch_numopt.curvature_estimator
  • torch_numopt.line_search
  • torch_numopt.numerical_optimizer
  • torch_numopt.objective
  • torch_numopt.solve_system
  • torch_numopt.step_initializer
  • torch_numopt.trust_region
  • torch_numopt package
  • torch_numopt.line_search

torch_numopt.line_search#

Line-search algorithms for step-length determination.

This module provides various line-search strategies: backtracking, interpolation, and bisection. Each solver implements a specific method and can be combined with different stopping conditions (Armijo, Wolfe, Goldstein, etc.).

Functions

create_line_search_solver(method, condition)

Factory function to instantiate a line-search solver.

Classes

BacktrackingLineSearch([condition, c1, c2, ...])

Backtracking line search with step reduction.

BisectionLineSearch([condition, c1, c2, ...])

Bisection (binary search) line search.

InterpolationLineSearch([condition, c1, c2, ...])

Line search using quadratic/cubic interpolation.

LineSearchSolver([condition, c1, c2, tau, ...])

Abstract base class for line-search solvers.

previous

torch_numopt.curvature_estimator

next

torch_numopt.numerical_optimizer

Show Source

© Copyright 2024, Eugenio Lorente-Ramos.

Created using Sphinx 8.1.3.

Built with the PyData Sphinx Theme 0.19.0.